Criminal Punishment Prediction Based on Fuzzy Document Vector Model

Abstract

The drug addict problem is more and more critical to the world during current periods, analysing the criminal process and predicting the final punishment from judgments report is an interesting and important task. Existing studies on text analysis and language model supply methods based on special feature selection and ontology models generation which need external knowledge by human experts. In this paper, however, we creatively leveraged such text data onto prediction in the public judgments without human business. We propose a combined framework to capture the prediction problem by considering both valued based rules and fuzzy document models. This framework contains the complete process as: information extraction, term fuzzy and document vector regression. We setup an experiment on a real-world dataset and compare our model with traditional classification and regression methods. The results show that our model outperforms than others by both RMSE and R squared measures.