A classification method to deal with category imbalance

Abstract

The invention discloses an accusation classification method for dealing with unbalanced classification, which comprises the following steps: acquiring and collecting corpus, preprocessing the case description corpus to obtain the case description corpus related to the accusation; The case description corpus related to a crime is taken as the positive example corpus, and the case description corpusirrelevant to a crime is taken as the negative example corpus, and the training corpus and the test corpus are divided. Under-sampling algorithm is used to extract a number of subsets from the negative instance corpus independently and randomly, and each subset and the positive instance sample are combined into a training corpus subset. Several LSTM-based base classifiers are trained by using several subsets of training corpus. Combined with the classification results of the base classifier, the new crime description is classified. The invention can train a base classifier with high classification accuracy rate under the condition that the number of positive samples is small and the number of negative samples is large, solves the situation that the classification error rate is high underthe condition of category imbalance, and realizes the automatic inference of charges of case description.

Description

technical field [0001] The invention relates to the technical field of natural language processing, in particular to a crime classification method for dealing with category imbalance. Background technique [0002] With the development of artificial intelligence technology, more and more traditional industries have introduced artificial intelligence technology to reduce manpower, save costs and improve efficiency. In the daily work of the court, it is necessary to infer the charge of a piece of case description for the convenience of recording and retrieval. The traditional method is mostly manual entry, which requires personnel with certain legal knowledge and a certain understanding of each crime. After reading the case description, manually enter the crimes related to the case into the system. There are many types of crimes. After working for a long time, the staff cannot complete the entry of crimes efficiently. Therefore, there is an urgent need for an automatic charge...

AI Technical Summary

Problems solved by technology

However, the classification effect of this fully supervised classification method is highly dependent on the quality of the human-labeled corpus, and the binary classifier often faces the situation that the number of negative samples is large and the number of positive samples is small

The imbalance of categories will cause the classification accuracy of the two classifiers to drop greatly, which will eventually affect the results of multi-classification

Images