Research on the Criminal Recidivism Prediction Based on Machine Learning Algorithm
Jiaxin Zhang · Atlantis Highlights in Computer Sciences/Atlantis highlights in computer sciences · 2023
Criminologists and social security personnel around the world have found that the risk of released criminals is much higher than that of people who have not committed crimes, and preventing people with criminal records from recommitting crimes should be one of the strategic priorities of social crime prevention.Therefore, risk assessment of criminal recidivism has been used to improve social security by predicting the criminal recidivism of offenders.In order to predict criminal recidivism, this article applied machine learning (ML) algorithms models (KNN, random forest, support vector machine and logistic regression) on the data set of the basic information about 10,000 criminal defendants in Broward County, Florida and their recidivism within two years.The predictive accuracy of models used in this article was between 0.64 and 0.67, with AUC ranging between 0.65-0.72.The AUC value of logistic regression is highest with 0.713 while support vector machine has the highest accuracy reaching to 0.671.This study provides a reference on selecting best method to predicting criminal recidivism.