Application of Clustering-based Entropy Weighted Association Analysis

Abstract

In recent years, Association analysis has been applied to various areas, such as disease diagnosis, risk management, bioinformatics, web mining, etc. However, the support-confidence framework of association rule mining has some limitations. For the analysis of complex Chinese medicine data, it will generate unreasonable and misleading results. For the accurate inference of relationships between symptoms and medications, we take different measures into account. We first classify measures by hierarchical clustering method. The measures are divided into four categories: {Gini index, AddedValue}, {Φ, maxconf, Jaccard, DCC}, {cosine, certainty} and {kulczynski, kappa, Laplace}.Then, we select various measures to construct the entropy-weighted association analysis model. The ROC curve is used to analyze the inference effect, and we compare the AUC with the Bayesian network. The results illustrate that the proposed method performs well.