Pattern Matching Strategy Based on Optimized Fuzzy Clustering Algorithm for Industrial Internet
Sheng Wang, Jinkuan Wang, Yinghua Han · 2019
With the promotion of Industrial Internet and big data technologies, pattern matching based on previous case database is widely employed. As the database growth and case libraries continue to scale, the conventional method can't be maintained at a good performance continuously. Thus, an improved method is proposed based on optimized Fuzzy c-means (FCM) algorithm in this paper. The historical case base is dimensioned and classified into each subclass set by clustering of FCM algorithm combining with principle components analysis (PCA) results, which is beneficial to eliminate the impact of redundant variables. In order to reduce the complexity of case retrieval, a graded matching strategy is proposed. In the primary matching stage, the target case is matched with centers obtained by clustering, and the secondary matching stage, the matching case for operational optimization is retrieved among the subclass set corresponding the center obtained in primary stage. Furthermore, a comprehensive metric is proposed to improve the measure of case similarity, which effectively utilizes Euclidean distance and Cosine similarity for nearest neighbor (NN) queries in high dimensional sample spaces. The experimental study is carried out with UCI data set, and shows that the method proposed in the paper can get better clustering results, which the matching time is shortened and the accuracy is increased.