Artificial Neural Network Prediction of Psychological Well-Being Among Adolescents Using Family, Peer, School, and Individual Factors

Abstract

Objective: This study aimed to develop and evaluate an artificial neural network model for predicting psychological well-being among adolescents using family, peer, school, and individual factors. Methods and Materials: This cross-sectional predictive study was conducted among 612 adolescents aged 13–18 years enrolled in secondary schools in Ontario, Canada, during the 2025–2026 academic year. Participants were selected using multistage sampling. Psychological well-being was assessed using the Warwick–Edinburgh Mental Well-Being Scale. Family-related predictors included family cohesion and family support; peer-related predictors included peer support and friendship quality; school-related predictors included school connectedness, academic engagement, and academic standing; and individual predictors included self-esteem, resilience, emotion regulation difficulties, perceived stress, and selected demographic characteristics. An artificial neural network based on a multilayer perceptron architecture was developed using 70% of the sample for training, 15% for validation, and 15% for independent testing. Model performance was evaluated using mean squared error, root mean squared error, mean absolute error, correlation coefficients, and coefficient of determination. Permutation-based feature importance was used to rank predictors. Findings: The artificial neural network demonstrated strong predictive performance, with R² values of .83, .76, and .74 in the training, validation, and test samples, respectively. In the independent test sample, the model achieved R = .86, RMSE = 4.44, MAE = 3.52, and MSE = 19.74. Perceived stress was the most influential predictor, followed by self-esteem, resilience, emotion regulation difficulties, family support, and school connectedness. Conclusion: Psychological well-being among adolescents can be predicted with substantial accuracy using a multidimensional artificial neural network model, with individual psychological resources and stress-related factors contributing most strongly while family and school contexts also providing meaningful predictive value.