Elena Kokthi1,2*†
Fatmir Guri1,2†- 1Department of Food Science and Biotechnology, Agriculture University of Tirana, Tirana, Albania
- 2Department of Economics and Policies of Rural Development, Agricultural University of Tirana, Tirana, Albania
This study explores the predictors of household food waste among young consumers (N = 414) in Albania, a country undergoing a transition from traditional to urbanized food systems. Using Principal Component Analysis (PCA), we identified eight dietary patterns, three waste patterns, three categories of reasons for food waste, and one dimension of dietary association. These components were analyzed alongside demographic characteristics through Random Forest Regression (RFR) and Artificial Neural Networks (ANN) to evaluate their predictive capacity. The results show that food waste is systematically linked to dietary regimes: perishable fresh foods are wasted due to storage and planning deficits, while protein and convenience-based diets drive waste through over-purchasing and portioning errors. Forest Regression (RFR) models consistently outperformed Artificial Neural Networks (ANNs) in predictive accuracy, with higher R2 values (0.47–0.62 vs. 0.15–0.37) and lower error rates, demonstrating the strength of combining PCA with ML techniques. The findings highlight the behavioral pathways behind waste and provide a novel approach to modeling sustainability challenges in transitioning food systems.
1 Introduction
Albania's food system is undergoing a rapid transition marked by urbanization, lifestyle change, and dietary modernization (United Nations Economic Commission for Europe, 2024; Vincze et al., 2023). Within this evolving context, household food waste represents a critical challenge to achieving Sustainable Development Goal 12.3 on responsible consumption and production (Manzoor et al., 2024; Sarangi et al., 2024). Although global estimates suggest that one-third of food produced for human consumption—around 1.3 billion tons—is lost or wasted each year (FAO, 2011), understanding where and why waste occurs at the household level is essential for designing context-specific sustainability policies. This study situates household waste within Albania's ongoing food system transition, where traditional Mediterranean diets increasingly coexist with convenience and ultra-processed foods, revealing how this coexistence reshapes consumption and waste behaviors.
Research distinguishes between developed and developing countries regarding where food losses occur along the supply chain (Adaryani et al., 2025; Rahman et al., 2024). In developed economies, 40–50% of total food waste occurs at the consumption stage—mainly in households and food service settings—driven by overpurchasing, portion sizes, aesthetic preferences, and poor planning (Baykoca and Yilmaz, 2025; Ishangulyyev et al., 2019). By contrast, in developing countries, losses occur primarily upstream during production, handling, and storage due to infrastructural and technological limitations (Cattaneo et al., 2021; FAO, 2011). These differences indicate that food waste is shaped not only by consumer behavior and cultural norms but also by systemic capacities in production and distribution, suggesting that mitigation strategies must be context-specific.
Recent studies have emphasized that food waste cannot be fully understood without considering the structure of diets and the behavioral mechanisms that accompany them (Birney et al., 2017; Franco et al., 2022; Grummon et al., 2023; Iori et al., 2022). Dietary patterns influence what is eaten and, indirectly, what is wasted: diets rich in perishable or convenient foods often lead to greater discard due to spoilage, over-purchasing, or inadequate storage, whereas health-oriented and plant-based diets may foster more mindful consumption (Iori et al., 2022; Tseng et al., 2022). These relationships are further shaped by cognitive and planning behaviors, such as food management, awareness, and purchasing routines, highlighting the importance of integrating what people eat with how they behave within a unified analytical framework (Adaryani et al., 2025; Roodhuyzen et al., 2017). However, this interrelation remains underexplored in transitioning economies such as Albania, where rapid dietary modernization reshapes both consumption and waste practices.
Young people in Albania—and across the Western Balkans—stand at the forefront of the ongoing nutrition transition, gradually moving from traditional diets toward convenience-oriented and ultra-processed foods (Gurazi et al., 2025; Kokthi et al., 2025a; Llanaj et al., 2018). Understanding how these evolving dietary practices and household food-management behaviors shape food waste is crucial for designing effective sustainability interventions.
This study applies a mixed analytical framework that integrates Principal Component Analysis (PCA) to identify dietary and waste typologies, and two machine-learning models—Random Forest Regression (RFR) and Artificial Neural Networks (ANN)—to predict food waste outcomes among young urban consumers in Albania. By combining behavioral, cognitive, and demographic factors, the study aims to reveal how individual consumption logics translate into waste generation patterns.
The research specifically addresses the following questions:
RQ1: How do dietary patterns, diet association, and household food-management behaviors interact with demographic factors to influence food waste among young urban consumers?
RQ2: Which machine-learning approach—Random Forest Regression (RFR) or Artificial Neural Networks (ANN)—provides stronger predictive accuracy and interpretative value for modeling food waste behavior?
2 Literature review
2.1 Dietary patterns, consumer behavior, and their implications for food waste
The relationship between dietary patterns, consumer behavior, and the transition toward sustainable diets has gained increasing attention in both public health and environmental research (Cattaneo et al., 2021; Chidi et al., 2022; Feng et al., 2020). Diets rich in plant-based foods, including fruits, vegetables, legumes, and whole grains, are associated with improved health outcomes and reduced environmental footprints (Auestad and Fulgoni, 2015; Nelson et al., 2016). These dietary patterns contribute to lower greenhouse gas emissions, reduced land use, and better overall population health, positioning them as key strategies in addressing both non-communicable diseases and ecological degradation (Auestad and Fulgoni, 2015; Hoek et al., 2017; Nelson et al., 2016; Sáez-Almendros et al., 2013).
Consumer behavior plays a pivotal role in shaping these dietary patterns (Arnold, 2022; dos Santos et al., 2022; Iori et al., 2022; Matthiessen et al., 2025). Socio-demographic characteristics such as age, gender, and education, together with cultural norms and attitudes toward sustainability, influence what people eat and how they manage food resources (dos Santos et al., 2022; Grasso et al., 2019; Iori et al., 2022). For instance, individuals who frequently consume organic products are more likely to maintain healthier diets and report lower rates of obesity, highlighting the interconnectedness between consumption habits, health awareness, and environmental values (Azzurra et al., 2019; Matthiessen et al., 2025; Saraiva et al., 2021).
However, while these associations are well-established, research has paid less attention to how what people eat influences what they waste. Dietary composition directly shapes food waste potential—plant-based diets may generate more avoidable waste due to perishability, whereas convenience-oriented diets increase waste through over-purchasing and poor planning. Therefore, understanding dietary patterns alongside behavioral mechanisms such as awareness, planning, and storage practices is essential for explaining household-level waste dynamics. In summary, dietary patterns determine not only the nutritional and environmental quality of diets but also the likelihood and nature of food waste. Accordingly, this study hypothesizes that dietary patterns significantly influence household food waste generation (H1).
2.2 Dietary profiles and the behavioral paradox of healthy food waste
Dietary patterns significantly influence both the type and quantity of food consumed and, consequently, the volume and nature of food waste generated. Different dietary profiles prioritize food categories with varying levels of perishability, storage demands, and spoilage risks (Dernini et al., 2016; dos Santos et al., 2022; Matthiessen et al., 2025; Ridoutt et al., 2017; Seed and Rocha, 2018). Plant-based, health-oriented diets rich in fruits, vegetables, dairy, and whole grains are frequently endorsed for their health and environmental benefits (Dernini et al., 2016; Knaapila et al., 2022). Nevertheless, these same items are among the most perishable, leading to disproportionately high levels of avoidable household food waste (Luu, 2020; Phooi et al., 2022). This contradiction, in which higher dietary quality does not necessarily lead to lower food waste, is often referred to as the healthy-diet paradox (Huang and Tseng, 2020; Messner et al., 2020; Porpino et al., 2015). It underscores the pivotal role of consumer behavior in determining whether healthy diets translate into environmentally sustainable practices (Saraiva et al., 2021). A nutritious food profile alone is insufficient unless coupled with intentional consumption behaviors (dos Santos et al., 2022). Evidence shows that households with high awareness of both health and sustainability issues are more likely to follow structured food routines, reduce impulsive purchases, and creatively repurpose leftovers, thereby mitigating over-preparation and spoilage (Pickering, 2023; Romani et al., 2018). In summary, although health-oriented diets are environmentally desirable, their potential benefits can be offset by behavioral inefficiencies in food management. This behavioral paradox highlights that dietary improvement must be accompanied by stronger awareness and planning practices to reduce waste. Accordingly, this study hypothesizes that healthy or plant-based dietary profiles are associated with higher waste potential when behavioral control is ineffective (H2).
2.3 Dietary profiles, demographics, and food waste
Demographic and contextual variables may moderate the diet–waste relationship. Income, household size, and place of residence influence both dietary adoption and food management capabilities (Grasso et al., 2019; Ilakovac et al., 2020). Higher-income households, for instance, may be better positioned to afford diverse, high-quality diets, but without supportive behaviors, they also tend to waste more (Grasso et al., 2019; Porpino et al., 2015). Likewise, large households often face logistical challenges in coordinating meals and portioning them appropriately. Evidence from China, Zhang et al. (2024) shows that households with higher dietary-preference scores—indicating a more deliberate focus on health and nutrition—waste less food, likely because they plan meals more effectively, use ingredients more diversely, and assign greater value to food.
Young adults, particularly young women who frequently manage household food provisioning and preparation, play a key role in household food-waste outcomes (Barak et al., 2023; Romani et al., 2018). Young adulthood (roughly ages 18–30) is a transitional life stage in which individuals take on independent food-management responsibilities (shopping, storage, cooking), often for the first time (Melnyk et al., 2025). These transitions shape persistent habits that affect waste generation later in life (Cantaragiu, 2019).
In many cultural contexts, women retain primary responsibility for household food work (grocery shopping, meal preparation, storage decisions; Gamhewage et al., 2015; Kokthi et al., 2025b), situating them as the principal agents whose practices determine avoidable household waste quantities (Janssens et al., 2019).
Across studies, several recurring determinants explain demographic variation in food-waste behavior:
1. Behavioral skills and food literacy (Janssens et al., 2019; Vásquez Neyra et al., 2022)
2. Planning and in-store purchase behavior (Iori et al., 2022; Janssens et al., 2019; Stancu et al., 2016)
3. Socioeconomic constraints and context—budget pressures can both reduce waste (via frugality) and increase it (bulk buying, poorly matched package sizes; Abeliotis et al., 2014; Urugo et al., 2024; Vásquez Neyra et al., 2022)
4. Household structure and responsibility clarity—shared housing or unclear distribution of foodwork (common among young adults in student/shared households), leads to duplicated purchases and abandoned items, increasing waste risk (Janssens et al., 2019) and
5. Gendered social norms and moral responsibility often lead women to internalize social expectations around provisioning and hospitality.
In summary, demographic characteristics condition how dietary choices are translated into food-management outcomes. Gender and age shape the behavioral routines through which food is purchased, prepared, and discarded. Consequently, the relationship between dietary patterns and food waste cannot be understood without considering these moderating demographic influences. Therefore, this study hypothesizes that demographic factors—specifically gender and age—moderate the relationship between dietary patterns and household food-waste behavior (H3).
2.4 Use of machine learning (ML) approaches in food waste prediction
Food waste and dietary behaviors are driven by complex, often non-linear interactions among individual behavior, demographics, supply-chain factors, and temporal/contextual variables.
Random Forest regression (RFR) is increasingly used in food waste analysis at the consumption stage due to its ability to handle complex, non-linear relationships among multiple factors influencing waste generation (Izquierdo-Horna et al., 2025; Nijloveanu et al., 2024). This machine learning approach builds numerous decision trees and averages their predictions, providing robust and accurate estimates of food waste based on diverse input variables such as consumer behavior, household characteristics, and regional indicators (Petrisia and Kularatne, 2024; Turker, 2025; Yang, 2024). Its strength lies in managing large and small datasets with missing values and revealing the relative importance of different predictors, enabling targeted interventions to reduce waste (Petrisia and Kularatne, 2024; Rodrigo et al., 2021). Similarly, artificial neural networks (ANNs) have been increasingly applied across the food system to capture the multifactorial, non-linear relationships and often outperform linear models for prediction and pattern discovery (Anggraeni et al., 2021; Bigné et al., 2010; Huang et al., 2007; Shi et al., 2024). ANNs approximate arbitrary non-linear mappings and can model high-dimensional interactions between behavioral constructs (e.g., attitudes, habits), objective inputs (e.g., household size), and outcomes (waste mass, composition, shelf life) without strong parametric assumptions (Huang et al., 2007; Shi et al., 2024). While most behavioral studies still rely on traditional multivariate techniques, ML techniques such as ANNs have been applied to integrate psychometric constructs (e.g., planning, label reading, storage skills) with demographic inputs to predict waste-avoidance behaviors observed during crises (e.g., COVID-19) and routine contexts (Vásquez Neyra et al., 2022).
Empirical comparisons show that ANN approaches frequently yield superior predictive accuracy compared to linear regression or simple time-series models in waste-generation forecasting and hospitality-sector waste estimation (Ali Abdoli et al., 2012; Azarmi et al., 2018).
Similarly, a recent systematic review on the prediction of consumer preferences using machine learning, including Random Forest (RF) and Artificial Neural Networks (ANN), provides meta-analytic insights into their comparative performance in consumer behavior studies (Byrne et al., 2022). According to Byrne et al. (2022), Deep Neural Networks (DNNs), a form of ANN, achieved the highest predictive accuracy (85–94%), followed by RFR algorithms with slightly lower accuracy (around 80%). The review emphasizes that while ANN models excel in capturing complex patterns and achieving higher classification accuracy in consumer preference prediction, RF models offer better interpretability and robustness, reducing the risk of overfitting. In the present study, we will utilize both to benefit from the strengths of each ML.
The interrelation among dietary patterns, reasons for waste, and demographic moderators is summarized in Figure 1, which illustrates the study's conceptual framework. The model integrates an analytical sequence—from input variables to PCA-based dimensional reduction and machine-learning modeling (RFR and ANN)—to balance interpretability and predictive accuracy in explaining food waste patterns.
Figure 1. Conceptual framework linking dietary patterns, reasons for food waste, and demographic moderators to food waste patterns through PCA and machine learning. Source: Author's elaboration.
3 Materials and methods
3.1 Data source and questionnaire
We adapted the questionnaire from Iori et al. (2022) for the Western Balkans context, applying forward–back translation and minor wording adjustments while retaining the constructs and response formats. Cognitive pre-testing (n =230) ensured semantic equivalence. Item sources and adaptations are structured into four main components: dietary consumption patterns, food waste behavior, self-perceived diet type, and reasons for food waste, along with demographic characteristics. Dietary patterns were assessed by asking respondents to indicate the frequency with which they consumed 41 food categories (ranging from cold meats, soft drinks, fruits, vegetables, dairy products, legumes, and sauces to pasta, rice, and bread). Response options followed an eight-point Likert-type scale, where 1 indicated “every day” and eight indicated “I do not know.” These items were later used to extract dietary components through Principal Component Analysis (PCA), and each was assigned a short variable name (e.g., cold meat, raw fish, yogurt) for ease of reference and coding.
The questionnaire also included a self-assessment item on dietary identity, asking respondents to define their current dietary approach, i.e., (i) vegan, (ii) vegetarian, (iii) healthy/low-fat, (iv) Mediterranean (characterized by pasta and pizza), (v) organic or sustainability-oriented, (vi) based on local products, (vii) traditional or national diets, (viii) smart or packaged-product oriented, (ix) undefined or (x) no specific preference, and “I do not know.” Each response was coded numerically (1–10) under the variable “diet define” and used as an indicator of food-related cognitive orientation.
Demographic information was also collected to assess the influence of socio-economic and household variables on food-related behavior. The variables included age, gender, place of residence, household size, family composition, presence of children in the household, education level, and employment status. These demographic characteristics were used as covariates in the analysis to account for heterogeneity in consumption and food waste behaviors across different population segments. All variables were pre-coded and harmonized for statistical analysis. The structured design of the questionnaire enabled the application of advanced multivariate techniques, including Principal Component Analysis (PCA) to identify latent dietary and waste typologies, and ML modeling to explore how dietary profiles, perceived dietary orientation, perceived reasons for food waste, and demographic features predict food waste behavior.
3.2 Sampling and data collection
The study employed a convenience sampling approach using an online questionnaire distributed across various Albanian regions, primarily through university networks and social media platforms targeting students and young professionals. A total of 420 valid responses were collected, with the majority residing in Tirana (68.6%), followed by Gjirokastra (10.0%), Kamez (4.8%), Durrës (3.8%), and other urban or semi-urban centers. This distribution mirrors Albania's demographic concentration, as Tirana—the country's educational and economic nucleus—naturally accounts for a large share of digitally active youth and higher-education students. The overrepresentation of this urban cohort is thus analytically justified, given the study's focus on urban food transitions and sustainability behaviors.
The sample composition provides valuable context for understanding the socio-demographic foundations of dietary and waste behaviors. Most respondents were young adults aged 18–24 years (60.9%), a group characterized by transitional consumption patterns, irregular meal planning, and a growing reliance on convenience foods—factors often associated with avoidable waste. Women accounted for 80.5% of the total sample, consistent with their traditionally central role in household food management and purchasing decisions, underscoring the importance of their inclusion for behavioral interpretation.
Household size averaged 4.38 members (SD = 1.43), with families of four to five members forming the majority. Larger households tend to purchase in greater volumes, increasing both the risk of overbuying and the opportunity for economies of scale in food utilization. Family composition adds further nuance: 24.3% were couples aged 35–54 with children, while 31% consisted of young adults living with peers or relatives—settings that illustrate the importance of coordination and shared responsibility in food management. Additionally, 35.7% of participants reported having children under 18, a demographic group often associated with portion-size variability and leftover generation (see Table 1).
Table 1. Descriptive statistics of the sample by demographic characteristics.
Educational attainment was notably high: 60.3% of respondents held a bachelor's degree and 26.6% a master's degree. This reflects the strong representation of university-affiliated participants and suggests a population with elevated awareness of sustainability and consumption-related issues. However, as confirmed by numerous studies (e.g., Tonini et al., 2023) such awareness does not automatically translate into reduced waste, highlighting a behavioral knowledge–action gap. Finally, 41.3% of respondents were employed full-time, a condition that often correlates with time scarcity and the adoption of convenience-oriented food practices, including pre-prepared meals and takeaway consumption—both of which are relevant to understanding waste generation in urban contexts.
3.3 Data analysis
The statistical analysis proceeded in two main stages. In the first stage, we applied Principal Component Analysis (PCA) with Varimax rotation to reduce the dimensionality of dietary intake, reasons for food waste, and waste. Sampling adequacy was assessed through the Kaiser–Meyer–Olkin (KMO) measure and Bartlett's Test of Sphericity, both of which indicated suitability for factor analysis. Components with eigenvalues greater than one were retained, and loadings above 0.50 were considered significant for interpretation. In the second stage, we employed machine learning approaches to predict waste using dietary patterns (DP), diet associations (DA), reasons for food waste (RFW), and demographic variables. Two algorithms were selected for their complementary strengths: Random Forest Regression (RFR) and Artificial Neural Networks (ANN). RFR is an ensemble decision tree method that aggregates multiple bootstrapped trees to improve predictive accuracy and reduce overfitting (Breiman, 2001). It provides measures of variable importance, making it particularly suitable for behavioral data where interpretability is essential. ANN, by contrast, is a non-linear modeling technique that simulates interconnected “neurons” across hidden layers to capture complex patterns and interactions (Baykoca and Yilmaz, 2025; Nijloveanu et al., 2024; Yang, 2024). It is less interpretable than RFR but offers advantages in reducing individual-level prediction error and modeling non-linear relationships that tree-based approaches may miss.
The dataset was randomly split into training, validation, and test sets, ensuring that model parameters were optimized on the training data, tuned on the validation data, and evaluated on the test data. The ANN architecture comprised multiple hidden layers and was tuned iteratively to minimize validation mean squared error. For RFR, hyperparameters such as the number of trees and the number of features per split were adjusted to minimize out-of-bag (OOB) error. Model performance was assessed using multiple metrics: the coefficient of determination (R2), root mean squared error (RMSE), mean absolute error (MAE), mean absolute percentage error (MAPE), and coefficient of variation of RMSE (CV-RMSE). This multi-metric evaluation provided a robust assessment of both explanatory power and predictive precision.
All analyses were conducted on the full sample (N = 414), which was predominantly comprised of young individuals (18–24 years), females, urban residents, and highly educated individuals. This demographic structure was particularly relevant for interpreting model outputs, as ANN tended to elevate the influence of demographics, while RFR consistently highlighted dietary and behavioral predictors. The combination of PCA with RFR and ANN will provide an interpretable framework for identifying structural drivers of food waste and a flexible modeling approach capable of capturing demographic distinctions. All statistical analyses were performed using a complementary software environment: IBM SPSS Statistics v26 was employed for descriptive statistics and preliminary diagnostics (KMO, Bartlett's test, Eigenvalues, and factor loadings), while JASP 0.9510 was used to conduct the Principal Component Analysis (PCA) and to implement the Machine-Learning models—Random Forest Regression (RFR) and Artificial Neural Network (ANN)—for predictive modeling and validation.
4 Empirical results
4.1 Self-defined diet types results
The results on self-perceived dietary patterns or diet association (DA) provide a meaningful lens into how individuals conceptualize their food practices (Figure 2). This subjective dimension reflects not only actual behavior but also personal values, aspirations, and identity. Among 414 respondents, the most common self-identified pattern was “healthy, low-fat” (30.0%), followed by “unspecified” (21.7%), “traditional” (17.4%), “Mediterranean” (10.1%), and “local products” (8.2%), as shown in Figure 2. Smaller shares selected organic/sustainable (4.3%), vegetarian (1.4%), vegan (1.0%), and SMART/packaged (0.5%), while 5.3% reported “do not know.” Nearly 27% of respondents did not specify or were unsure of their dietary style, a lack of clarity that may signal reduced agency in everyday food choices; such uncertainty can contribute to inefficient planning and, ultimately, higher levels of food waste.
Figure 2. Distribution of self-defined dietary patterns (N = 414). Source: Author's elaboration.
4.2 Dietary patterns based on reported consumption frequencies
Respondents reported high-frequency consumption of staple foods, including fresh bread, milk and yogurt, cheese, fresh vegetables, and root vegetables, with a majority consuming them daily or several times a week. Notably, fresh fruit had the highest daily consumption (56.7%), aligning with nutritional guidelines and showing a promising orientation toward health-conscious eating.
Foods such as cold meats, soft drinks, fast food, and ready or semi-ready meals were reported with moderate frequency (mostly 2–3 times per week). Products such as alcohol, raw red and white meat, raw fish, and pre-cooked/frozen meals were generally consumed rarely or sporadically. The low intake of raw meats and frozen products may reflect cultural cooking practices that favor fresh preparation, as well as limited access or affordability. Interestingly, alcohol showed the highest proportion of “rare” responses (32.4%), suggesting low regular consumption in this cohort, though underreporting due to social desirability bias cannot be ruled out.
Cooked meats, both white and red, were consumed 2–3 times per week, indicating a moderate intake of animal-based protein. The relatively low prevalence of vegetarian or vegan diets, as shown above confirms that plant-based alternatives have not yet reached mainstream status.
Moderate consumption of dipping sauces, cooking sauces, and mayonnaise suggests a preference for enhanced flavors and processed condiments, which may increase overall sodium and fat intake. These items are typically underexamined in dietary surveys, yet they play a significant role in total dietary quality (Fardet et al., 2024). Overall, the dietary patterns observed in the sample suggest a hybrid consumption model, in which traditional food staples coexist with the growing adoption of convenience and processed products.
4.3 Food waste patterns
The analysis of food waste quantities reported over the past 7 days reveals significant patterns that reflect the vulnerability of specific food categories and broader inefficiencies in household food management. The most frequently wasted items were perishable foods, including fresh fruits, vegetables, dairy products, cooked and raw meats, and bread, with most of the waste reported in this category.
From a frequency perspective, 57.6% of respondents report wasting food “rarely” or “less than once a week,” reflecting a promising baseline of awareness and cost-conscious behavior. However, the 29% of individuals who report wasting food 1–2 times weekly, and the 9% who admit to wasting food daily, signal persistent gaps in behavioral practices and food literacy. These findings emphasize that small, frequent waste outcomes, although perceived as negligible by individuals, accumulate into significant food waste over time. The data show that food waste in Albanian households is both widespread and concentrated in key food categories, particularly perishables and the most frequently consumed items.
4.4 Principal component analysis (PCA) on dietary patterns
To identify the underlying dimensions of food consumption behavior, a PCA was performed on 41 food items reported in the questionnaire. The analysis yielded eight components with eigenvalues greater than 1, cumulatively explaining 69.6% of the variance. The Kaiser-Meyer-Olkin (KMO) measure was 0.920, indicating excellent sampling adequacy, and Bartlett's Test of Sphericity was significant (Chi-Square = 12687.21, df = 820, p < 0.001), confirming the suitability of the data for PCA. The extracted components were interpreted and named based on the dominant food items with loadings greater than 0.5; see Table 2 for the names of dietary patterns and their respective loadings.
Table 2. Dietary patterns from PCA (rotated component matrix).
4.5 Principal component analysis (PCA) on food waste patterns
The PCA on food waste behavior identified three main components, explaining 74.1% of the total variance, indicating a robust structure of waste typologies (Table 3). The first component, labeled “Meat and Convenience Waste,” grouped high-waste items such as cooked and raw meats (red and white), pizza, sausages, and sandwiches, foods that are typically expensive, perishable, and consumed in social or fast-paced contexts. The second component, Dairy and Plant-Based Waste, comprises items such as yogurt, milk, cheese, fresh vegetables, and fish, which are associated with health-oriented diets but prone to spoilage. The third component, Cold and Packaged Waste, includes soft drinks, packaged bread, frozen foods, and breakfast products. This pattern aligns with recent literature on the role of Westernized diets and ultra-processed food consumption in shifting food waste profiles (Aschemann-Witzel et al., 2017; Frehner et al., 2021). However, some studies often overlook these long-shelf-life products, assuming they are less likely to be wasted. The data presented in this study challenge that assumption, suggesting that such products may still be wasted due to consumer neglect, storage overflow, or a lack of awareness, especially among younger or urban populations (Lehmann, 2011).
Table 3. PCA results on food waste dimensions: factor composition and variance explained.
4.6 Principal component analysis (PCA) on reasons for food waste
The PCA on self-reported reasons for food waste confirmed sampling adequacy, with a Kaiser–Meyer–Olkin (KMO) value of 0.825, exceeding the recommended threshold of 0.6. In addition, Bartlett's Test of Sphericity was significant (χ2 = 1268.115, df = 78, p < 0.001), indicating sufficient correlations among the variables for PCA. The analysis extracted three components after Varimax rotation, which together explained 52% of the variance: The first component, Over-Purchasing & Planning Deficit, included inability to estimate needs (Cant_Estimate, 0.733), purchases due to sales (Buy_Sales, 0.717), buying food not liked (Buy_NotLike, 0.662), shopping rarely (Shop_Rarely, 0.646), fear of shortage (Fear_Shortage, 0.626), forgetting food (Forget_Food, 0.564), and buying large amounts (Buy_Large, 0.500). The second component, Inefficient Storage & Preparation, was defined by poor storage (Poor_Storage, 0.688), not reusing leftovers (No_Leftovers, 0.689), bulk buying (Buy_Bulk, 0.624), and cooking too much (Cook_TooMuch, 0.574). The third component, Expiration & Spoilage, was narrower but precise, with fruit spoilage (Fruit_Spoil, 0.750) and food expiring (Food_Expiring, 0.657).
These results suggest that household food waste in the sample is driven by a combination of planning and purchasing deficits, inefficient household management, and constraints related to perishability.
4.7 Correlations between dietary patterns, waste patterns, and reasons for waste
The correlation analysis revealed several significant associations between dietary patterns, food waste patterns, and self-reported reasons for waste. Several statistically significant relationships emerged, underscoring the interplay between consumption typologies and waste generation across the surveyed young urban population.
Dietary Pattern 1 (Protein-Oriented Traditional Diet) demonstrated the strongest positive associations with Protein & Convenience Waste (WP1; r = 0.189, p < 0.01) and Staple & Condiment Waste (WP3; r = 0.249, p < 0.01), as well as with Over-Purchasing & Planning Deficit (RFW1; r = 0.244, p < 0.01). This pattern typifies households that preserve traditional meat-based habits while adapting to convenience-oriented purchasing, a behavioral convergence often linked to over-preparation and inadequate portion control.
Dietary Pattern 2 (Fresh & Dairy-Rich Mediterranean Diet) displayed negative correlations with Staple & Condiment Waste (WP3; r = −0.159, p < 0.01) and Over-Purchasing & Planning Deficit (RFW1; r = −0.138, p < 0.01), suggesting that health-conscious consumers embedded in Mediterranean dietary norms exhibit stronger planning routines and a higher sense of food value, leading to reduced avoidable waste.
Dietary Pattern 3 (Fast-Food & Convenience Diet) correlated positively with Protein & Convenience Waste (WP1; r = 0.241, p < 0.01) and Staple & Condiment Waste (WP3; r = 0.136, p < 0.01), reflecting the structural inefficiencies inherent in convenience consumption—namely, packaging redundancy, impulse buying, and low meal-preparation literacy.
Dietary Pattern 4 (Alcohol & Cooked Meat Diet) was positively associated with Staple & Condiment Waste (WP3; r = 0.159, p < 0.01), Over-Purchasing & Planning Deficit (RFW1; r = 0.172, p < 0.01), and Expiration & Spoilage (RFW3; r = 0.110, p < 0.05). The confluence of these associations suggests that sporadic, socialized consumption patterns generate irregular leftovers and poorly synchronized storage habits.
Dietary Pattern 5 (Sweet & Spread-Based Diet), though not statistically strong, showed a weak positive correlation with Perishable Fresh Waste (WP2; r = 0.080, p = 0.102) and Expiration & Spoilage (RFW3; r = 0.073, p = 0.163), implying that discretionary foods—often purchased for hedonic gratification—carry latent waste risks through neglect rather than deliberate disposal.
Dietary Pattern 6 (Staples & Canned Foods Diet) correlated positively with Perishable Fresh Waste (WP2; r = 0.122, p < 0.05) and Staple & Condiment Waste (WP3; r = 0.227, p < 0.01), but was negatively associated with Over-Purchasing & Planning Deficit (RFW1; r = −0.029, n.s.), reflecting the behavioral paradox of long-shelf-life foods—low planning failures but high accumulation leading to expiration.
Dietary Pattern 7 (Protein-Balanced Diet) maintained near-zero correlations across all waste and behavioral dimensions (r = −0.025–0.094), underscoring its statistical neutrality and behavioral equilibrium within the dataset.
Finally, Dietary Pattern 8 (Carbohydrate-Rich Diet) exhibited a dual profile: negative associations with Protein & Convenience Waste (WP1; r = −0.097, p < 0.05) and Perishable Fresh Waste (WP2; r = −0.099, p < 0.05), but a positive relationship with Staple & Condiment Waste (WP3; r = 0.148, p < 0.01), suggesting that carbohydrate-heavy diets may buffer short-term waste through routine consumption while still generating losses from surplus pantry goods.
Collectively, these findings illustrate that dietary typologies structure waste generation through distinct behavioral mechanisms. Patterns rich in perishable or convenience products (DP2, DP3, DP4) predispose consumers to specific waste profiles, whereas traditional and staple-oriented diets (DP1, DP6) exhibit structural inefficiencies linked to portioning, storage, and purchasing inertia.
The subsequent mapping of waste factors to dietary groups provides the analytical foundation for the machine-learning phase. Specifically, Protein & Convenience Waste (WP1) was driven by fast-food, alcohol, and meat-heavy diets (DP3, DP4, DP7), reflecting behavioral impulsivity and irregular planning. Perishable Fresh Food Waste (WP2) aligned with fresh and carbohydrate-rich diets (DP2, DP8), where spoilage stems from short shelf life rather than indifference. Staple & Condiment Waste (WP3) was linked to sweet/spread-based and staple-heavy diets (DP5, DP6), indicating that even durable food categories contribute to waste when subject to habitual overstocking.
This integrated pattern supports the application of Random Forest Regression (RFR) and Artificial Neural Networks (ANN) to quantify predictive strength and interaction complexity. RFR isolates variable importance with interpretive transparency, while ANN captures non-linear relationships, jointly providing a behavioral map of how dietary diversity, storage practices, and planning deficits converge to explain household food waste in transitioning urban food systems.
4.8 Results on food waste through ML
Following the correlation analysis, the next stage applies machine-learning models to uncover the predictive structure underlying the observed relationships. By integrating Random Forest Regression and Artificial Neural Networks, the analysis moves from statistical association to behavioral prediction, capturing the combined influence of dietary patterns, behavioral reasons for waste, and demographics. The following subsections present these predictive results across the three identified waste patterns, beginning with Protein & Convenience Waste (WP1).
4.8.1 Protein & convenience waste: predictive modeling through random forest regression
The Random Forest model achieved a satisfactory predictive performance for Protein & Convenience Waste, with an R2 of 0.576, RMSE of 0.738, MAE of 0.639, and MAPE of 76.32%. The relatively low mean squared error and the positive R2 indicate a moderate fit, suggesting that the model captured meaningful relationships between dietary patterns, reasons for waste, and socio-demographic predictors. The importance analysis revealed that the strongest predictors were RFW1 (Over-purchasing & Planning Deficit), RFW3 (Expiration & Spoilage), and DP3 (Fast-Food & Convenience Diet). These findings indicate that households with higher reliance on convenience foods and weaker planning behaviors are more prone to wasting protein- and convenience-related food categories. Demographic variables such as family composition, dietary association (DA), and education also contributed, but with lower relative importance compared to dietary and waste-related factors.
4.8.2 Protein & convenience waste: predictive modeling through an artificial neural network
The ANN model achieved lower explanatory power than RFR, with R2 = 0.352, RMSE = 0.857, MAE = 0.622, and a high MAPE of 116.67%. Despite a weaker overall performance, the ANN highlighted additional nuances in predictor importance. The most influential variables were again DP3 (Fast-Food & Convenience Diet) and DA (Dietary Association), followed by education, household children, and employment status. The ANN placed greater weight on demographic and household-level factors than RFR, capturing non-linear interactions that may explain heterogeneity across households. The dominance of DP3 across both models reinforces the central role of convenience-oriented diets in driving waste within this category.
RFR and ANN results consistently indicate that protein and convenience waste are strongly associated with convenience-driven dietary practices and planning-related waste behaviors (RFW1, RFW3). The RFR provided better predictive accuracy, while the ANN revealed more nuanced contributions of demographic variables, particularly the presence of children in the household and employment status. These complementary insights suggest that interventions to improve food planning and awareness of convenience diets could substantially reduce protein- and convenience-related food waste. Diagnostic figures for both models are presented in Figure A2 (RFR) and Figure A3 (ANN).
4.8.3 Perishable fresh food waste: predictive modeling through random forest regression
The RFR model for Perishable Fresh Food Waste explained a moderate share of variance (R2 = 0.470), with RMSE = 0.898, MAE = 0.699, and MAPE = 702.98%. Although predictive accuracy was lower compared to Waste Pattern 1, the results highlight meaningful associations between fresh food waste and both dietary and behavioral factors. Feature importance analysis identified DP2 (Fresh & Dairy-Rich Diet) and DP8 (Carbohydrate-Rich Diet) as the most influential dietary predictors, alongside RFW2 (Inefficient Storage & Preparation) and RFW1 (Over-Purchasing & Planning Deficit). These results suggest that households with strong adherence to Mediterranean-style diets rich in perishable products but with inadequate storage and preparation practices are more likely to waste fruits, vegetables, and dairy products. Demographic predictors, such as household size, family composition, and age, also played a role, indicating that larger or younger households may be more susceptible to the challenges of fresh food preservation.
4.8.4 Perishable fresh food waste: predictive modeling through artificial neural network (ANN)
The ANN model achieved a weaker fit, with R2 = 0.148, RMSE = 1.120, MAE = 0.850, and MAPE = 356.79%. Despite its lower predictive performance relative to RFR, the ANN provided complementary insights into variable importance. It highlighted education, household size, DP8 (Carbohydrate-Rich Diet), and age as dominant factors, alongside RFW1 and RFW3 (Expiration & Spoilage). Compared to RFR, the ANN placed greater emphasis on demographic and structural variables, such as household composition and education, reflecting the non-linear role of these factors in moderating fresh food spoilage risks.
The combined evidence suggests that Perishable Fresh Food Waste is primarily driven by dietary reliance on fresh and perishable foods (DP2, DP8) and by inefficiencies in household management (RFW2, RFW1). The RFR model provided stronger predictive power, while the ANN provided additional evidence that household-level variables, such as education and size, critically shape waste outcomes. These findings emphasize the dual importance of technical skills (e.g., storage and preparation) and socio-demographic context in reducing fresh food waste. Figures for both models are provided in Figure A4 (RFR) and Figure A5 (ANN).
4.8.5 Staples & condiment waste: predictive modeling through random forest regression
The RFR model performed strongly for Staples & Condiment Waste, achieving an R2 of 0.615, RMSE of 0.649, MAE of 0.569, and MAPE of 101.47%. This indicates the highest predictive accuracy among the three waste patterns, reflecting the relatively structured, less perishable nature of staple products. Feature importance results identified DP6 (Staples & Canned Foods Diet) and DP5 (Sweet & Spread-Based Diet) as the most influential dietary predictors, alongside RFW1 (Over-Purchasing & Planning Deficit) and RFW3 (Expiration & Spoilage). These findings suggest that waste in this category is primarily explained by poor planning and over-purchasing behaviors, which increase the likelihood of storing large quantities of durable foods that eventually expire or remain unused. Demographic variables, such as age, family composition, and household size, also played a significant role, highlighting differences in consumption and storage practices across household types.
4.8.6 Staples & condiment waste: predictive modeling through artificial neural network
The ANN model yielded a moderate predictive capacity, with R2 = 0.365, RMSE = 0.897, MAE = 0.644, and MAPE = 113.64%. Feature importance analysis showed a slightly different emphasis compared to RFR: RFW2 (Inefficient Storage & Preparation) emerged as the most influential predictor, followed by education, RFW1, employment, and DA (Dietary Association). Dietary patterns DP6 and DP5 remained significant but ranked lower than in RFR. This shift highlights how ANN captured non-linear interactions between waste behaviors and socio-demographics, underscoring the role of household capacity and education in mediating the extent of waste in non-perishable product categories.
The combined results suggest that Staples & Condiment Waste is driven by behavioral inefficiencies (RFW1, RFW2, RFW3) in interaction with staple-oriented diets (DP6, DP5). RFR provided the highest overall predictive accuracy, while ANN provided additional insight into the role of household education and employment dynamics. This indicates that durable food waste is not only a result of poor planning and over-purchasing but also of inefficient storage and household management practices. Figures for both models are presented in the Figure A6 (RFR) and Figure A7 (ANN).
Across all three waste patterns, the machine learning analyses confirmed that dietary patterns and household waste behaviors (RFW1–RFW3) are the strongest predictors of food waste. At the same time, socio-demographic variables provide important but secondary explanatory power. The Random Forest Regression (RFR) models consistently outperformed the Artificial Neural Networks (ANN) in predictive accuracy, with higher R2 values (0.47–0.62 vs. 0.15–0.37) and lower error rates. RFR excelled in identifying the central role of specific dietary profiles—particularly DP3 (Fast-Food & Convenience) for protein- and convenience-related waste, DP2 (Fresh & Dairy-Rich) and DP8 (Carbohydrate-Rich) for perishable waste, and DP6 (Staples & Canned Foods) for durable food categories. By contrast, the ANN models, while less accurate, offered complementary insights by emphasizing the non-linear influence of demographic and household factors, such as education, household size, and the number of children in the household.
Overall, the results highlight that RFR is more reliable for predictive modeling in this context. In contrast, ANNs contribute interpretive depth, particularly regarding the interplay between household structure and socio-economic conditions. Taken together, the findings underscore that effective interventions must address both diet-driven consumption styles and household-level management practices, while also considering the demographic realities that shape waste behaviors across different food categories.
5 Discussion of results
This study provides a multi-layered analysis of household food waste by combining PCA with ML approaches (Random Forest Regression and Artificial Neural Networks). By examining eight dietary patterns, three waste patterns, three behavioral reasons for waste, and key demographic factors, the results yield a nuanced understanding of how urban, young, and predominantly female populations in a transitioning food system navigate food consumption and waste in dense urban environments. Rather than viewing waste as an outcome of isolated factors, the findings demonstrate a systemic relationship between what is eaten, how it is managed, and who consumes it—a pattern that both confirms and extends previous research on food waste behavior.
5.1 Dietary patterns and their role in waste generation
The eight dietary patterns identified through PCA mirror typologies observed in European and Mediterranean contexts, where traditional, Westernized, and hybrid diets coexist (Bôto et al., 2022; Vásquez Neyra et al., 2022). This confirms earlier work suggesting that food systems in transition often reflect a blend of modernization and tradition (Dernini et al., 2016). However, this study advances previous findings by empirically linking each dietary pattern to specific waste profiles through predictive modeling.
For instance, the Fast-Food & Convenience Diet (DP3) was the strongest predictor of Protein & Convenience Waste (WP1), reinforcing prior evidence that convenience-driven food choices increase discard rates of meat and ready-to-eat products (Franco et al., 2022; Stancu et al., 2016). Similarly, the Fresh & Dairy-Rich Mediterranean Diet (DP2) and Carbohydrate-Rich Diet (DP8) were associated with Perishable Fresh Food Waste (WP2), aligning with Cao and Li (2023), who noted that the high perishability of healthy foods leads to unintended waste. Notably, the association between Staples & Canned Foods (DP6) and Staples & Condiment Waste (WP3) challenges the assumption that durable foods are waste-resilient, showing that even low-perishability items are vulnerable when planning is poor. Thus, this study contributes by revealing that sustainability-oriented diets do not automatically ensure low waste unless paired with effective management practices.
5.2 Behavioral reasons for waste
The behavioral dimensions extracted from PCA—Over-Purchasing & Planning Deficit (RFW1), Inefficient Storage & Preparation (RFW2), and Expiration & Spoilage (RFW3)—clarify how cognitive and routine-based mechanisms interact with dietary choices. These results are consistent with behavioral economics research showing that impulsive purchasing and weak planning drive most avoidable household waste (Pickering, 2023; Romani et al., 2018). However, the current findings extend this evidence by specifying how diet type moderates these behavioral tendencies: convenience diets exacerbate impulsive buying (RFW1), Mediterranean diets amplify storage-related waste (RFW2), and staple-based diets accumulate expiration-related waste (RFW3). This multi-level linkage between food types and behavioral inefficiencies adds explanatory depth beyond previous single-variable models (Secondi et al., 2015; Stancu et al., 2016), suggesting that interventions must target both the content of consumption and the behavioral process simultaneously.
5.3 Machine learning insights
Machine Learning methods provided complementary insights into the complexity of food waste determinants. The Random Forest Regression (RFR) model offered the most accurate predictions (R2 up to 0.62 for WP3), confirming the dominant role of dietary and behavioral predictors. In contrast, the Artificial Neural Network (ANN) model captured subtler non-linear effects related to socio-demographics. This duality demonstrates that food waste behavior is not merely additive but emerges from interacting variables and threshold effects.
Compared with prior studies that employed traditional linear regressions (Secondi et al., 2015) the ML approach reveals previously hidden heterogeneity—especially among education levels, household structures, and employment categories. ANN's ability to detect non-linear dynamics highlights that even small behavioral or contextual shifts can disproportionately alter waste outcomes. Hence, combining interpretable (RFR) and adaptive (ANN) models provides both explanatory clarity and behavioral sensitivity, underscoring the methodological value of ML in sustainability behavior analysis (Chan et al., 2025; Izquierdo-Horna et al., 2025; Nijloveanu et al., 2024).
5.4 Demographic influences
Demographic effects observed here echo findings in both developed and emerging economies (Aschemann-Witzel, 2016; Ishangulyyev et al., 2019; Principato et al., 2015) but also reveal new nuances. The predominance of young, urban females confirms that these cohorts are central to both dietary modernization and waste generation (Gamhewage et al., 2015). Consistent with earlier evidence, higher education and employment status correlate with improved awareness, yet greater exposure to convenience culture (Porpino et al., 2015).
These results suggest a dual role of youth: agents of change who value sustainability but face behavioral and structural constraints—such as time scarcity, inadequate storage, and information overload—that maintain waste. The Albanian context extends existing literature by illustrating how post-transition urbanization fosters both opportunity and risk: Mediterranean-oriented diets support health but, without behavioral adaptation, lead to high spoilage rates. This aligns with Bravi et al. (2020), who document similar convergence in dietary practices between the Balkans and Western Europe.
5.5 Implications for policy and practice
The findings have several implications. First, interventions should not only promote healthier dietary patterns but also address the management challenges these diets create, particularly with perishable foods. Second, policy campaigns targeting youth and students could focus on meal planning, portion control, and storage practices, leveraging digital tools and nudges to improve behavior. Third, the importance of waste from staples and condiments suggests that interventions should extend beyond fresh food, incorporating awareness of dry goods and processed items often wasted due to poor estimation. Ultimately, urban planning and retail strategies should consider the impact of time scarcity and convenience culture, which contribute to waste among working urban populations. These measures align with integrated approaches to sustainable consumption that address both what people buy and how they handle it.
5.6 Limitations and future research
Despite its contributions, the study has limitations. First, the reliance on self-reported data may introduce recall and social desirability biases. Second, while ANN was applied to model non-linearities, its performance was constrained by the sample size. Larger and more diverse datasets would likely improve predictive accuracy. Third, the cross-sectional design limits causal inference, and longitudinal studies could better capture how life-cycle stages and evolving dietary practices affect waste.
Future research should build on the combined PCA–ML framework, applying it across countries to explore cultural variability in food waste determinants. Furthermore, integrating transactional data (e.g., grocery receipts) or sensor-based measures (e.g., smart bins) could enhance predictive accuracy and reduce reliance on self-reports.
5.7 Theoretical implications
This study advances theoretical understanding in two ways. First, it extends the behavioral–dietary linkage framework by empirically demonstrating that food waste emerges from the intersection of diet structure, behavioral routines, and demographic context. The “healthy-diet paradox” identified here supports the view that sustainability transitions depend on behavioral competence as much as on dietary composition. Second, the use of ML methods adds conceptual depth beyond conventional linear modeling. Machine Learning captures behavioral complexity—non-linear thresholds, feedback loops, and interactions—revealing that minor behavioral deviations (e.g., planning gaps or storage inefficiencies) can trigger disproportionate waste outcomes. These complements established psychological and economic models by revealing multidimensional, context-dependent patterns that traditional regressions overlook. In this sense, ML does not merely enhance prediction accuracy but broadens theoretical insight into the adaptive and heterogeneous nature of sustainable consumption behavior.
6 Conclusions
This study indicates that food waste is not a random by-product of consumption, but a predictable outcome of the interaction between dietary patterns, self-defined diet associations, reasons for waste, and waste patterns, all embedded in the demographic context of a young, urban, and predominantly female population. By applying a mixed-methodological framework that integrates Principal Component Analysis (PCA), Random Forest Regression (RFR), and Artificial Neural Networks (ANN), we provide evidence on how food choices and household management behaviors translate into specific categories of food waste. The 8×3×3×1 structure allowed us to capture the multidimensionality of food waste, linking what people eat, how they identify their diets, why food is wasted, and which food categories are most likely to be discarded.
Methodologically, the dual PCA–ML approach demonstrates the value of combining interpretable and adaptive models to capture both linear and non-linear relationships underlying sustainable consumption behavior.
From a policy perspective, the results highlight the need to address food management behaviors alongside the promotion of healthy diets. While Mediterranean and plant-based diets support sustainability goals, their high perishability requires complementary interventions in planning, storage, and portion management. Educational campaigns targeting youth and working urban populations should leverage digital tools and behavioral nudges to encourage mindful food practices. At a systemic level, integrating waste-reduction principles into urban retail design—through smaller packaging, flexible portions, and time-saving options—can support behaviorally compatible sustainability transitions.
Future research should expand this framework cross-nationally to explore cultural variability in the diet–waste nexus and integrate objective data sources such as smart-bin tracking or purchase receipts to enhance accuracy. By linking behavioral economics with advanced analytics, this study contributes to a more nuanced theoretical understanding of sustainable consumption—demonstrating that behavioral adaptation, rather than dietary change alone, is pivotal for reducing household food waste and achieving long-term food system resilience.
Data availability statement
The raw data supporting the conclusions of this article will be made available by the authors, without undue reservation.
Ethics statement
Written informed consent was obtained from the individual(s) for the publication of any potentially identifiable images or data included in this article.
Author contributions
EK: Conceptualization, Formal analysis, Methodology, Software, Writing – original draft, Writing – review & editing. FG: Conceptualization, Supervision, Validation, Writing – review & editing.
Funding
The author(s) declare that financial support was received for the research and/or publication of this article. This work was partly funded by the National Agency for Scientific Research and Innovation (NASRI) in Albania.
Acknowledgments
My sincere gratitude to the generous support of the National Agency for Scientific Research and Innovation (NASRI) in Albania, which enabled us to carry out this study. The financial support received for our project “Measuring the Socio-Economic Impact of Women as Key Actors in Poverty Reduction and the Sustainable Development of Food Systems in Albania”, based on Decision No. 10, date 15.08.2023, “On the approval of the financing of winning projects of the National Research and Development Program for the Period 2023–2024”, is responsible for the significant success of the study.
Conflict of interest
The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.
Generative AI statement
The author(s) declare that Gen AI was used in the creation of this manuscript. Generative AI is used for English proofreading.
Any alternative text (alt text) provided alongside figures in this article has been generated by Frontiers with the support of artificial intelligence and reasonable efforts have been made to ensure accuracy, including review by the authors wherever possible. If you identify any issues, please contact us.
Publisher's note
All claims expressed in this article are solely those of the authors and do not necessarily represent those of their affiliated organizations, or those of the publisher, the editors and the reviewers. Any product that may be evaluated in this article, or claim that may be made by its manufacturer, is not guaranteed or endorsed by the publisher.
Supplementary material
The Supplementary Material for this article can be found online at: https://www.frontiersin.org/articles/10.3389/frsus.2025.1717100/full#supplementary-material
References
Abeliotis, K., Lasaridi, K., and Chroni, C. (2014). Attitudes and behaviour of Greek households regarding food waste prevention. Waste Manage. Res. J. Sustain. Circular Econ. 32, 237–240. doi: 10.1177/0734242X14521681
Adaryani, R. L., Palouj, M., Gholami, H., Baghestany, A. A., Damirchi, M. J., Dadar, M., et al. (2025). Predicting household food waste behavior: bringing food literacy and purchasing power into the theory of planned behavior. J. Retail. Consumer Services 82:104119. doi: 10.1016/j.jretconser.2024.104119
Ali Abdoli, M., Falah Nezhad, M., Salehi Sede, R., and Behboudian, S. (2012). Longterm forecasting of solid waste generation by the artificial neural networks. Environ. Progress Sustain. Energy 31, 628–636. doi: 10.1002/ep.10591
Anggraeni, M. C., Silaban, C. A., Anggreainy, M. S., and Cahyadi, E. (2021). “Role of artificial intelligence in the management of food waste,” in 2021 2nd International Conference on Artificial Intelligence and Data Sciences (AiDAS) (Malaysia), 1–6. doi: 10.1109/AiDAS53897.2021.9574167
Arnold, N. (2022). Standards and waste: valuing food waste in consumer markets. Worldwide Waste 5:2. doi: 10.5334/wwwj.84
Aschemann-Witzel, J. (2016). Waste not, want not, emit less. Science 352, 408–409. doi: 10.1126/science.aaf2978
Aschemann-Witzel, J., Jensen, J. H., Jensen, M. H., and Kulikovskaja, V. (2017). Consumer behaviour towards price-reduced suboptimal foods in the supermarket and the relation to food waste in households. Appetite 116, 246–258. doi: 10.1016/j.appet.2017.05.013
Auestad, N., and Fulgoni, V. L. (2015). What current literature tells us about sustainable diets: emerging research linking dietary patterns, environmental sustainability, and economics. Adv. Nutr. 6, 19–36. doi: 10.3945/an.114.005694
Azarmi, S. L., Oladipo, A. A., Vaziri, R., and Alipour, H. (2018). Comparative modelling and artificial neural network inspired prediction of waste generation rates of hospitality industry: the case of North Cyprus. Sustainability 10:2965. doi: 10.3390/su10092965
Azzurra, A., Massimiliano, A., and Angela, M. (2019). Measuring sustainable food consumption: a case study on organic food. Sustain. Prod. Consum. 17, 95–107. doi: 10.1016/j.spc.2018.09.007
Barak, F., Efitre, J., Odong, R., and Melgar-Quiñonez, H. (2023). Women's agency in nutrition in the association between women's empowerment in agriculture and food security: a case study from Uganda. World Food Policy 9, 228–249. doi: 10.1002/wfp2.12063
Baykoca, B., and Yilmaz, S. (2025). Understanding food loss patterns across developed and developing countries using a GDP, growth rate, and health expenditure-based typology. Sci. Rep. 15:27597. doi: 10.1038/s41598-025-13156-3
Bigné, E., Aldas-Manzano, J., Küster, I., and Vila, N. (2010). Mature market segmentation: a comparison of artificial neural networks and traditional methods. Neural Comput. Appl. 19, 1–11. doi: 10.1007/s00521-008-0226-y
Birney, C. I., Franklin, K. F., Davidson, F. T., and Webber, M. E. (2017). An assessment of individual foodprints attributed to diets and food waste in the United States. Environ. Res. Lett. 12:105008. doi: 10.1088/1748-9326/aa8494
Bôto J. M. Rocha A. Miguéis V. Meireles M. Neto B. (2022). Sustainability dimensions of the Mediterranean diet: a systematic review of the indicators used and its results. Adv. Nutr. 13, 2015–2038. doi: 10.1093/advances/nmac066
Bravi, L., Francioni, B., Murmura, F., and Savelli, E. (2020). Factors affecting household food waste among young consumers and actions to prevent it. A comparison among UK, Spain and Italy. Resour. Conserv. Recycl. 153:104586. doi: 10.1016/j.resconrec.2019.104586
Byrne, A., Bonfiglio, E., Rigby, C., and Edelstyn, N. (2022). A systematic review of the prediction of consumer preference using EEG measures and machine-learning in neuromarketing research. Brain Inf. 9:27. doi: 10.1186/s40708-022-00175-3
Cantaragiu, R. (2019). The impact of gender on food waste at the consumer level. Studia Universitatis Vasile Goldis Arad Econ. Ser. 29, 41–57. doi: 10.2478/sues-2019-0017
Cao, X., and Li, X. (2023). Will healthy diets increase household food waste? Front. Sustain. Food Syst. 7:1180741. doi: 10.3389/fsufs.2023.1180741
Cattaneo, A., Sánchez, M. V., Torero, M., and Vos, R. (2021). Reducing food loss and waste: five challenges for policy and research. Food Policy 98:101974. doi: 10.1016/j.foodpol.2020.101974
Chan, J., Cao, K., Fong, L. H. N., and Nong, S. Z. (2025). Predicting food leftovers in shared dining: a machine learning approach. Int. J. Contemp. Hospital. Manage. 37, 3304–3326. doi: 10.1108/IJCHM-02-2025-0283
Chidi, B. S., Okudoh, V. I., Hutchinson, U. F., Ngongang, M. M., Maphanga, T., Madonsela, B. S., et al. (2022). A perspective on emerging inter-disciplinary solutions for the sustainable management of food waste. Appl. Sci. 12:11399. doi: 10.3390/app122211399
Dernini, S., Berry, E. H., Serra-Majem, L., Vecchia, C. L., Capone, R., Medina, F. X., et al. (2016). Med Diet 4.0: the Mediterranean diet with four sustainable benefits. Public Health Nutr. 20, 1322–1330. doi: 10.1017/S1368980016003177
dos Santos, J. I. A. S., da Silveira, D. S., da Costa, M. F., and Duarte, R. B. (2022). Consumer behaviour in relation to food waste: a systematic literature review. Br. Food J. 124, 4420–4439. doi: 10.1108/BFJ-09-2021-1075
Fardet, A., Gold, S., Delgado, A., Kopsahelis, N., Kachrimanidou, V., Kaur, L., et al. (2024). How can food processing achieve food and nutrition security? Sustain. Dev. 32:sd.2899. doi: 10.1002/sd.2899
Feng, B., van Dam, K. H., Guo, M., Shah, N., Passmore, S., and Wang, X. (2020). Planning of food-energy-water-waste (FEW2) nexus for sustainable development. BMC Chem. Eng. 2:4. doi: 10.1186/s42480-020-0027-3
Franco, S., Barbanera, M., Moscetti, R., Cicatiello, C., Secondi, L., and Massantini, R. (2022). Overnutrition is a significant component of food waste and has a large environmental impact. Sci. Rep. 12:8166. doi: 10.1038/s41598-022-11813-5
Frehner, A., De Boer, I. J. M., Muller, A., Van Zanten, H. H. E., and Schader, C. (2021). Consumer strategies towards a more sustainable food system: Insights from Switzerland. Am. J. Clin. Nutr. 115, 1039–1047. doi: 10.1093/ajcn/nqab401
Gamhewage, M. I., Sivashankar, P., Mahaliyanaarachchi, R. P., Wijeratne, A. W., and Hettiarachchi, I. C. (2015). Women participation in urban agriculture and its influence on family economy—Sri Lankan experience. J. Agric. Sci. Sri Lanka 10, 192–206. doi: 10.4038/jas.v10i3.8072
Grasso, A. C., Olthof, M. R., Boevé, A. J., van Dooren, C., Lähteenmäki, L., and Brouwer, I. A. (2019). Socio-demographic predictors of food waste behavior in Denmark and Spain. Sustainability 11:3244. doi: 10.3390/su11123244
Grummon, A. H., Lee, C. J. Y., Robinson, T. N., Rimm, E. B., and Rose, D. (2023). Simple dietary substitutions can reduce carbon footprints and improve dietary quality across diverse segments of the US population. Nat. Food 4, 966–977. doi: 10.1038/s43016-023-00864-0
Gurazi, V., Zejnelhoxha, S., Sulenji, M., Koxha, L., Protoduari, H., Arapi, K., et al. (2025). Cardiometabolic phenotypes and dietary patterns in Albanian University-enrolled young adults: cross-sectional findings from the nutrition synergies WHO-aligned sentinel platform. Nutrients 17:3395. doi: 10.3390/nu17213395
Hoek, A. C., Pearson, D., James, S. W., Lawrence, M. A., and Friel, S. (2017). Healthy and environmentally sustainable food choices: consumer responses to point-of-purchase actions. Food Qual. Prefer. 58, 94–106. doi: 10.1016/j.foodqual.2016.12.008
Huang, C.-H., and Tseng, H.-Y. (2020). An exploratory study of consumer food waste attitudes, social norms, behavioral intentions, and restaurant plate waste behaviors in Taiwan. Sustainability 12:22. doi: 10.3390/su12229784
Huang, Y., Kangas, L. J., and Rasco, B. A. (2007). Applications of artificial neural networks (ANNs) in food science. Crit. Rev. Food Sci. Nutr. 47, 113–126. doi: 10.1080/10408390600626453
Ilakovac, B., Voca, N., Pezo, L., and Cerjak, M. (2020). Quantification and determination of household food waste and its relation to sociodemographic characteristics in Croatia. Waste Manage. 102, 231–240. doi: 10.1016/j.wasman.2019.10.042
Iori, E., Masotti, M., Falasconi, L., Risso, E., Segrè, A., and Vittuari, M. (2022). Tell me what you waste and I'll tell you who you are. An eight-country comparison of consumers food waste related habits. doi: 10.20944/preprints202210.0445.v1
Ishangulyyev, R., Kim, S., and Lee, S. (2019). Understanding food loss and waste—why are we losing and wasting food? Foods 8:297. doi: 10.3390/foods8080297
Izquierdo-Horna, L., Kahhat, R., and Vázquez-Rowe, I. (2025). Applying random forest to forecast municipal solid waste generation from household fuel consumption. Resour. Conserv. Recycl. Adv. 27:200264. doi: 10.1016/j.rcradv.2025.200264
Janssens, K., Lambrechts, W., van Osch, A., and Semeijn, J. (2019). How consumer behavior in daily food provisioning affects food waste at household level in the Netherlands. Foods 8:10. doi: 10.3390/foods8100428
Knaapila, A., Michel, F., Jouppila, K., Sontag-Strohm, T., and Piironen, V. (2022). Millennials' consumption of and attitudes toward meat and plant-based meat alternatives by consumer segment in Finland. Foods 11:3. doi: 10.3390/foods11030456
Kokthi, E., Cazzetta, D., Berberi, E., and Hoxha, F. (2025a). “Modelling food waste: a recursive partitioning approach to consumer segmentation,” in Paper presented at 6th International Conference on Biosystems and Food Engineering, Budapest, Hungary.
Kokthi, E., Guri, F., and Dafku, Z. (2025b). Climate change awareness and urban food choices: exploring motivations for short food chain engagement. Urban Sci. 9:142. doi: 10.3390/urbansci9050142
Lehmann, S. (2011). Optimizing urban material flows and waste streams in urban development through principles of zero waste and sustainable consumption. Sustainability 3, 155–183. doi: 10.3390/su3010155
Llanaj, E., Ádány, R., Lachat, C., and D'Haese, M. (2018). Examining food intake and eating out of home patterns among university students. PLoS ONE 13:e0197874. doi: 10.1371/journal.pone.0197874
Luu, T. T. (2020). Reducing food waste behavior among hospitality employees through communication: dual mediation paths. Int. J. Contemporary Hospital. Manage. 32, 1881–1904. doi: 10.1108/IJCHM-09-2019-0779
Manzoor, S., Fayaz, U., Dar, A. H., Dash, K. K., Shams, R., Bashir, I., et al. (2024). Sustainable development goals through reducing food loss and food waste: a comprehensive review. Future Foods 9:100362. doi: 10.1016/j.fufo.2024.100362
Matthiessen, L. E., Smed, S., Jensen, J. D., Srednicka-Tober, D., Rossi, L., and Bügel, S. G. (2025). Eating green in Copenhagen: organic consumers' path to less meat and minimal food waste. Front. Sustain. Food Syst. 9:1534068. doi: 10.3389/fsufs.2025.1534068
Melnyk, D., Salem, M., Ertz, M., and Wagner, R. (2025). Being the “better” student: Intentions to reduce food waste. Discover Sustain. 6:201. doi: 10.1007/s43621-025-01021-2
Messner, R., Richards, C., and Johnson, H. (2020). The “prevention paradox”: food waste prevention and the quandary of systemic surplus production. Agric. Human Values 37, 805–817. doi: 10.1007/s10460-019-10014-7
Nelson, M. E., Hamm, M. W., Hu, F. B., Abrams, S. A., and Griffin, T. S. (2016). Alignment of healthy dietary patterns and environmental sustainability: a systematic review. Adv. Nutr. 7, 1005–1025. doi: 10.3945/an.116.012567
Nijloveanu, D., Ti?a, V., Bold, N., Popescu, D. A., Smedescu, D., Smedescu, C., et al. (2024). The development of a prediction model related to food loss and waste in consumer segments of agrifood chain using machine learning methods. Agriculture 14:1837. doi: 10.3390/agriculture14101837
Petrisia, N. T., and Kularatne, I. (2024). A waste reduction journey: data-driven and machine learning on perishable food waste optimisation in New Zealand retail supermarkets. Rere whio J. Appl. Res. Prac. 4, 22–31. doi: 10.34074/rere.00403
Phooi, C. L., Azman, E. A., Ismail, R., Arif Shah, J., and Koay, E. S. R. (2022). Food waste behaviour and awareness of Malaysian. Scientifica 2022, 1–11. doi: 10.1155/2022/6729248
Pickering, J. (2023). Household meal planning as anticipatory practice: the role of anticipation in managing domestic food consumption and waste. Geoforum 144:103791. doi: 10.1016/j.geoforum.2023.103791
Porpino, G., Parente, J., and Wansink, B. (2015). Food waste paradox: antecedents of food disposal in low income households. Int. J. Consum. Stud. 39, 619–629. doi: 10.1111/ijcs.12207
Principato, L., Secondi, L., and Pratesi, C. A. (2015). Reducing food waste: an investigation on the behaviour of Italian youths. Br. Food J. 117, 731–748. doi: 10.1108/BFJ-10-2013-0314
Rahman, T., deb, N., Alam, M. Z., Moniruzzaman, M., Miah, M. S., Horaira, M. A., et al. (2024). Navigating the contemporary landscape of food waste management in developing countries: a comprehensive overview and prospective analysis. Heliyon 10:e33218. doi: 10.1016/j.heliyon.2024.e33218
Ridoutt, B. G., Hendrie, G. A., and Noakes, M. (2017). Dietary strategies to reduce environmental impact: a critical review of the evidence base. Adv. Nutr. 8, 933–946. doi: 10.3945/an.117.016691
Rodrigo, H., Beukes, E. W., Andersson, G., and Manchaiah, V. (2021). Exploratory data mining techniques (decision tree models) for examining the impact of internet-based cognitive behavioral therapy for tinnitus: machine learning approach. J. Med. Internet Res. 23:e28999. doi: 10.2196/28999
Romani, S., Grappi, S., Bagozzi, R. P., and Barone, A. M. (2018). Domestic food practices: a study of food management behaviors and the role of food preparation planning in reducing waste. Appetite 121, 215–227. doi: 10.1016/j.appet.2017.11.093
Roodhuyzen, D. M. A., Luning, P. A., Fogliano, V., and Steenbekkers, L. P. A. (2017). Putting together the puzzle of consumer food waste: Towards an integral perspective. Trends Food Sci. Technol. 68, 37–50. doi: 10.1016/j.tifs.2017.07.009
Sáez-Almendros, S., Obrador, B., Bach-Faig, A., and Serra-Majem, L. (2013). Environmental footprints of Mediterranean versus Western dietary patterns: beyond the health benefits of the Mediterranean diet. Environ. Health 12:118. doi: 10.1186/1476-069X-12-118
Saraiva, A., Fernandes, E., and von Schwedler, M. (2021). The pro-environmental consumer discourse: a political perspective on organic food consumption. Int. J. Consum. Stud. 45, 188–204. doi: 10.1111/ijcs.12611
Sarangi, P. K., Pal, P., Singh, A. K., Sahoo, U. K., and Prus, P. (2024). Food waste to food security: transition from bioresources to sustainability. Resources 13:164. doi: 10.3390/resources13120164
Secondi, L., Principato, L., and Laureti, T. (2015). Household food waste behaviour in EU-27 countries: a multilevel analysis. Food Policy 56, 25–40. doi: 10.1016/j.foodpol.2015.07.007
Seed, B., and Rocha, C. (2018). Can we eat our way to a healthy and ecologically sustainable food system? Can. Food Stud. La Revue Canadienne Des Études Sur l'alimentation 5, 182–207. doi: 10.15353/cfs-rcea.v5i3.258
Shi, C., Zhao, Z., Jia, Z., Hou, M., Yang, X., Ying, X., et al. (2024). Artificial neural network-based shelf life prediction approach in the food storage process: a review. Crit. Rev. Food Sci. Nutr. 64, 12009–12024. doi: 10.1080/10408398.2023.2245899
Stancu, V., Haugaard, P., and Lähteenmäki, L. (2016). Determinants of consumer food waste behaviour: two routes to food waste. Appetite 96, 7–17. doi: 10.1016/j.appet.2015.08.025
Tonini, P., Odina, P. M., and Durany, X. G. (2023). Predicting food waste in households with children: Socio-economic and food-related behavior factors. Front. Nutr. 10:1249310. doi: 10.3389/fnut.2023.1249310
Tseng, C., Chetthamrongchai, P., Mahmudiono, T., Sharma, S. K., Al-Awsi, G. R. L., Abed, S. A., et al. (2022). Sustainability assessment of food industry with the approach of water, energy and food nexus. Food Sci. Technol. 42:e37922. doi: 10.1590/fst.37922
Turker, G. F. (2025). Reducing food waste in campus dining: a data-driven approach to demand prediction and sustainability. Sustainability 17:379. doi: 10.3390/su17020379
United Nations Economic Commission for Europe (2024). Resilient Urban Food Ecosystems: Food Map of Tirana – Food Resource Assessment. United Nations, 79. Available online at: https://unece.org/sites/default/files/2025-01/Tirana%20Food%20Map_0_0.pdf (Accessed September, 15, 2025).
Urugo, M. M., Teka, T. A., Gemede, H. F., Mersha, S., Tessema, A., Woldemariam, H. W., et al. (2024). A comprehensive review of current approaches on food waste reduction strategies. Comprehens. Rev. Food Sci. Food Saf. 23:e70011. doi: 10.1111/1541-4337.70011
Vásquez Neyra, J. M., Cequea, M. M., Gomes Haensel Schmitt, V., and Ferasso, M. (2022). Food consumption and food waste behaviour in households in the context of the COVID-19 pandemic. Br. Food J. 124, 4477–4495. doi: 10.1108/BFJ-07-2021-0798
Vincze, F., Muka, T., Eichelmann, F., and Llanaj, E. (2023). Eating out intensity, ultra-processed foods and BMI among Albanian youth. Public Health Nutr. 26, 2953–2962. doi: 10.1017/S1368980023002173
Yang, Y. (2024). Leveraging machine learning for food waste reduction: an analysis of predictive models. Appl. Comput. Eng. 112, 154–160. doi: 10.54254/2755-2721/2025.18115
Keywords: dietary patterns, food waste, Artificial Neural Network (ANN), Random Forest Regression (RFR), PCA, sustainable consumption behavior
Citation: Kokthi E and Guri F (2025) A PCA–machine learning framework for understanding household food waste: evidence from young urban consumers in Albania. Front. Sustain. 6:1717100. doi: 10.3389/frsus.2025.1717100
Received: 07 October 2025; Revised: 19 November 2025;
Accepted: 25 November 2025; Published: 18 December 2025.
Edited by:
Myriam Ertz, Université du Québec à Chicoutimi, CanadaReviewed by:
Mehrdad Kordi, Université du Québec à Chicoutimi, CanadaAws Horrich, Université du Québec à Chicoutimi, Canada
Copyright © 2025 Kokthi and Guri. This is an open-access article distributed under the terms of the Creative Commons Attribution License (CC BY). The use, distribution or reproduction in other forums is permitted, provided the original author(s) and the copyright owner(s) are credited and that the original publication in this journal is cited, in accordance with accepted academic practice. No use, distribution or reproduction is permitted which does not comply with these terms.
*Correspondence: Elena Kokthi, ZWtva3RoaUB1YnQuZWR1LmFs
†ORCID: Elena Kokthi orcid.org/0000-0002-2227-6820
Fatmir Guri orcid.org/0000-0002-9685-5024