A coal mine safety evaluation method based on concept drifting data stream classification

Abstract

Monitoring data in coal mine is essentially data stream. With the change of environment, coal mine monitoring data stream implied concept drifts. Coal mine safety evaluation can be seen as concept drifting data stream classification. The method proposed in this paper is based on random decision tree model, and it uses Hoeffding Bounds inequality and information entropy instead of random selection to determine the split point, and it uses the threshold determined by Hoeffding Bounds inequality detect concept drift. Experimental results show the method can better detect concept drifts in data stream, and it has better classification accuracy for data stream, and it provides a new practical approach for coal mine safety evaluation.