Evaluation of a predictive model of home environment factors for school-age children's experiences of abuse: Utilizing the Panel Study on Korean Children and machine learning techniques

Objectives The purpose of this study was to evaluate the performance of a prediction model to determine whether a child's home environment factors accurately predict the occurrence of physical and emotional abuse incidents by parents one year later and to identify important predictive factors. Methods The research data used the 11th and 12th data of the Panel Study on Korean Children. The input variables are demographic variables, parenting behavior and mental health-related variables reported by parents, and family relationship variables reported by children from the 12th data of 1,352 11-year-old children (664 girls and 668 boys). The output variable is the frequency of physical and mental abuse incidents reported by children who turned 11 in the 12th data. The XGBoost algorithm was used to develop the prediction model, and the prediction performance was compared by referring to the logistic regression model. Results First, the predictive performance of XGBoost models predicting the occurrence of child abuse cases was excellent, ranging from .97 to 1.00, and all were significantly better than the logistic regression model. Second, the most important variables predicting the occurrence of child abuse were the decrease in parental authoritative parenting behavior and the increase in authoritarian parenting behavior. In addition, Increased marital conflict and decreased household income were included as predictors of child abuse. Conclusions The child abuse prediction model developed using the Panel Study on Korean Children data and machine learning is an excellent model that accurately predicts the occurrence of abuse cases in advance, and shows the importance of prioritizing parental support.