Multivariate Feature Extraction and Radial Basis Function Neural Network for Predictive Modeling of Psychological Health States

Abstract

This study proposes a predictive model integrating factor analysis and a Radial Basis Function (RBF) neural network to assess college students’ psychological health. Data from 100 students were analyzed using factor analysis to extract 18 secondary indicators, which served as inputs for the RBF neural network. The dataset was split into 85% for training and 15% for validation. Performance metrics included a Karl Pearson correlation coefficient (R ≥ 0.85), root mean square error (RMSE ≤ 0.39), and classification accuracy for four health status categories (excellent, good, fair, poor). Statistical significance was confirmed via t-tests (p < 0.05), demonstrating the model’s reliability in predicting psychological health states with high accuracy and efficiency.