Application of Improved DBSCAN Clustering Algorithm on Industrial Fault Text Data

Abstract

The industrial fault text data are the special type of short texts, and they come from the records of faults in the factory. Clustering the industrial fault text data can reduce the redundant data and find out the hidden information, which is of great significance to improve the utilization of the industrial fault text data. The industrial fault text data are unstructured and irregular, so the clustering faces quite a few challenges. This paper introduces some existing algorithms for the clustering of short texts, and the shortcomings of them are briefly analyzed. This paper indicates that the main problem of the clustering of the industrial fault text data is the contradiction between the requirements and the setup of parameters, and it leads to low accuracy when cluster the corpus of different sizes. To increase the accuracy of clustering, an improved clustering algorithm is proposed which can solve this contradiction. The results of the comparative experiments show that the improved clustering algorithm has better performance than DBSCAN in corpus of different sizes on the industrial fault text data.