Integrating Classification, Regression, and Time Series Models to Assess Biochar Safety, Optimize Pollutant Removal, and Predict Environmental Impacts

Shreyashi Deb RoyGanesh Khekare^*Sejal ChhajedAdrine Sharon Victor

• Vellore Institute of Technology, Vellore, Vellore, India

Biochar, which is a high-carbon biomass pyrolysis byproduct, has considerable potential in environmental remediation, serving as a soil conditioner, a carbon sequestration substrate, and a wastewater treatment agent. Nevertheless, for its effective and safe application, thorough assessment techniques must be employed to analyze and measure the presence of potential risks like organic pollutants, metallic toxicants, and volatile organic compounds (VOCs). This research presents an automated framework based on artificial intelligence (AI), designed to evaluate the quality of biochar in real-time and enhance its environmental sustainability. The proposed system leverages data from the publicly available database to create biochar safety models for prediction. The system consists of three separate models: a classification model to evaluate the safety of biochar according to its chemical makeup, a regression model to estimate quantified levels of heavy metals, and a time series model to predict VOC emissions under different environmental conditions, facilitating evaluation of potential air quality effects. Performance results show that the Random Forest Regression model achieved a low Mean Squared Error (MSE) of 0.0046 and a strong R2 score of 0.9549, indicating high reliability in predicting heavy metal content, while the Random Forest Classifier achieved an external validation accuracy of 96.7%. The efficacy of the LSTM-based time series model in real-time environmental monitoring was demonstrated by the Mean Absolute Percentage Error (MAPE) accuracy of 87.14% in predicting VOC emissions. The multi-model system permits ongoing, precise monitoring while drastically minimizing human interaction and related errors. The AI models developed show great efficacy in classifying biochar safety, estimating the content of heavy metals, and estimating VOC emissions at future times. The system improves evaluation accuracy, operational efficiency, and production optimization while reducing disposal expenses and environmental hazards. This study provides a new contribution by integrating classification, regression, and time series analysis in one automated quality assessment system for biochar. It presents a scalable and smart solution that can be applied across environmental and industrial applications, enabling the wider integration of AI technologies into sustainable material management and environmental monitoring.