Using Genetic Algorithm to Augment Test Data for Penalty Prediction

摘要

With the development of smart court construction, a deep learning method has been introduced into the field of penalty prediction based on judicial text. Since the increasing parameters of the penalty prediction model, the size of the data set to test the performance of the model is gradually expanding. First, we use the data augmentation method to make some changes to the original data to obtain a large number of augmented data with the same label. Then, we use the multi-objective genetic algorithm to search for high-quality test data from a large number of augmented data, so as to improve the diversity of augmented data. Finally, we perform experiments. The results of actual judicial cases show that compared with the random method, augmented test data based on the genetic algorithm can better test the performance of the penalty prediction model.