In this study, first, we constructed an automated machine learning evaluation system using data from an online questionnaire survey conducted on all elementary and junior high school students in multiple municipalities to measure cognitive emotional behavior risk (moral disengagement, self-serving cognitive distortion, normative beliefs about aggression, CU traits, low self-control, and sensation seeking) and interpersonal environmental factors (parenting and domestic violence, friendship, teacher's leadership, and neighbors' collective efficacy). We then attempted to verify the prediction of student guidance cases involving problem behaviors, school absenteeism, and bullying. Second, we examined whether the predictive power of machine learning could be improved by adding indicators of social information processing using VR and game-version neuropsychological tests (Iowa gambling task, facial recognition task, Go/No Go task) to the measurement of cognitive emotional behavior risk factors. In Study 1, we attempted to construct a machine learning system using data from two surveys of approximately 20,000 students collected in 2020. Four types of machine learning algorithms (random forest, LightGBM, SVM, and logistic regression) were applied, and precision, recall, and the evaluation index F-measure were compared. Focusing on recall, we confirmed that logistic regression had a recall rate of approximately 70%, and that it had a higher precision rate and F-measure than other algorithms. In Study 2, we attempted to build a machine learning system using data from approximately 9,000 children and students collected in 2022, which added VR social information processing measurements and neuropsychological tests to the indicators from Study 1. We confirmed that logistic regression had a recall rate of approximately 90%, dramatically improving predictive power. Based on these findings, we demonstrated that it is possible to build an automatic evaluation system using the machine learning method "logistic regression" and to predict and evaluate the occurrence of student guidance incidents probabilistically.