AUTHOR=Qasrawi Radwan , Vicuna Polo Stephanny , Abu Khader Rami , Abu Al-Halawa Diala , Hallaq Sameh , Abu Halaweh Nael , Abdeen Ziad TITLE=Machine learning techniques for identifying mental health risk factor associated with schoolchildren cognitive ability living in politically violent environments JOURNAL=Frontiers in Psychiatry VOLUME=Volume 14 - 2023 YEAR=2023 URL=https://www.frontiersin.org/journals/psychiatry/articles/10.3389/fpsyt.2023.1071622 DOI=10.3389/fpsyt.2023.1071622 ISSN=1664-0640 ABSTRACT=Cognitive abilities have been evidenced to be affected by children’s health (physical and mental), nutrition, lifestyle, academic performance, and social support. However, little is known about the interrelated associated factors, including those most directly correlated to cognitive development. In this paper, we sought to measure the performance of machine learning models in the identification and prediction of the key multidimensional variables associated with schoolchildren’s cognitive development. The study used a data set of 4762 students aged 10-15 years old attending public and United Nations’ Relief Works Agency schools in Palestine. The data set was composed of 47 features related to cognitive ability, sociodemographic, mental health, physical health, nutrition, lifestyle, social support, and academic performance scores. Cognitive ability was measured using the Cognitive Standard Test. Six machine learning algorithms (Gradient Boosting, Support Vector Machine, Random Forest, Artificial Neural Network, Naïve Bayes, and k-nearest neighbors) were examined. The results indicated that the cognitive ability distribution among participants was: 24.3%, 49.7%, and 25% for low, average, and above average cognitive levels respectively. The machine learning models showed a high level of performance measure in predicting cognitive ability from the associated combined features. Gradient Boosting and Support Vector Machine reported the highest prediction accuracy (85%), while the other models reported acceptable prediction accuracies of over 80%. The results indicated that academic performance, smoking, post-traumatic stress disorder, gender, and school violence are the five highest associated factors with cognitive ability scores. Furthermore, the study found a strong association between phosphorus, zinc, vitamin C, iron, and folate intake with cognitive abilities. Machine learning techniques were found to be an effective tool in predicting cognitive abilities from the associated variables. The ML models show to be an effective tool for facilitating intervention and prevention programs that might enhance schoolchildren’s cognitive skills development.