Exploring the Legal Accuracy of Mathematical Models in Crime Sentencing Prediction in the Face of Criminal Procedure Law Practice Needs

Abstract

Abstract The rapid development of computer science and big data technology, as well as the practical needs of criminal procedure law, provides the possibility and opportunity for mathematical model prediction to intervene in the study of crime sentencing in criminal procedure law. In this paper, we design the legal knowledge base for criminal proceedings with reference to the relevant sentencing standards and sentencing elements in the criminal procedure law. The BERT model mines the key plot elements in the crime instrument, and the Sigmoid activation function is used to classify the key plot elements. Then, a two-layer linear regression model with constraints is introduced to predict the sentencing term of crimes. According to the results of sentencing prediction for crimes with various sentencing terms, the prediction model has a high legal accuracy, with the highest prediction accuracy (0.85) for the 0-3 month type of sentence. It was also found that the prediction model achieved better results in the prediction of sentencing for different types of offenses, with more accurate sentencing for the minor injury type of offenses, with an accuracy rate higher than 0.95. The sentencing prediction model in this paper can assist judges in handling sentencing cases and likewise lay the foundation for intelligent crime sentencing in criminal proceedings.