An Ensemble Classification Algorithm for Text Data Stream based on Feature Selection and Topic Model

How to mine valuable information that users are interested in from a continuous text data stream, text data stream classification has received widespread attention as a core technology to solve the problem. This paper proposes a text data stream ensemble classification algorithm that combines feature selection and topic model. Firstly, the mutual information feature selection method is used to remove features that are not related to classification. Secondly, the LDA topic model is used to establish the document-topic distribution. Finally, the pre-processed text data stream is classified by an ensemble classification model. The experimental results show that the proposed text data stream ensemble classification algorithm can improve the classification performance of text data stream.