Recidivism prediction model based on logistic regression

Using logistic regression to deal with the recidivism binary classification problem, and they are used factors as the characteristics to affecting reoffending establish the potential between factors and reoffendingmodel relationship, and then to comprehensively assess the risk of reoffending. Theory and statistical tests are used to screen the variables, construct and train the logistic regression model, and evaluate the optimization with accuracy, recall, F1 score, and AUC. to ensure reliable model performance. By analyzing the shortcomings of data dependence and linear limitation, the constructed recidivism prediction model has theoretically enriched the cross-research results of criminology and statistics, and in practice has provided the judiciary with scientific and effective evaluation tools, which helps to make accurate decisions and rationally allocate resources, and plays an indispensable role in maintaining social stability and reducing recidivism, and continues to promote the development of recidivism prediction in the direction of intellectualization and precision. It has a non-negligible impact on maintaining social stability and reducing recidivism, and continues to promote the development of recidivism prediction in the direction of intellectualisation and precision.