Research on Prisoner Psychological Symptoms Quick Screening Model Based on Ensemble Learning

Abstract

With the rapid development of information technology such as artificial intelligence and big data, the organic combination of these new technologies with traditional psychological research paradigms can effectively improve the research logic, research methods and research tools of traditional psychological measurement, improve the objectivity, accuracy and efficiency of traditional psychological measurement, and thus improve the limitations of traditional psychological evaluation methods. Based on the big data of 25214 community correctional prisoners SCL-90 symptom self-assessment scale samples in a province, this paper first uses the machine learning XGB algorithm to generate the importance ranking of the items (features) of the self-assessment scale, carry out dimension reduction processing and feature selection, and then constructs a fusion algorithm model for classification prediction. This model takes GBDT, RF, AdaBoost as the baseline model, and uses Voting algorithm for fusion processing, In order to avoid the error bias caused by a single model, through performance comparison and analysis, the accuracy of the fusion processing results is the highest, reaching 0.974, and the recall and F1 score are also the highest, reaching 0.90 and 0.92 respectively.