Application of Bagging Ensemble Classifier based on Genetic Algorithm in the Text Classification of Railway Fault Hazards

Abstract

The safety accident hidden danger of on-site inspection by railway workers are stored in text format, and this kind of data contains a lot of valuable information related to railway safety, so it is urgent to classify and manage the data by classification model. In this paper, we analyze the characteristics of such data. Firstly, we use TF-IDF method to extract text features and convert them into vectors. Then, decision tree classifier is used to classify the data. In order to improve classification accuracy, the Bagging Ensemble Classifier conduct a random sample training to text vector converted by TF-IDF which decision as the base classifier, produce Bagging classification results, considering the Bagging Algorithm is a the number of base classifiers results voting combination classification model which has a better classification performance, we use Genetic Algorithm to calculate Bagging classifier combination optimization, better classification results are produced by the Ensemble Classifier based on Genetic Algorithm(Evolutionary Ensemble Classifier). Through the experimental analysis of text data on safety accident hidden danger of power supply catenary in a Railway Bureau, it is proved that the safety classification accuracy, recall rate and f-score value of the Evolutionary Ensemble Classifier model are significantly improved.