Generation Method of Association Rules by Symmetric Hellinger Measure

Interest in big data has been growing in recent years. Data mining is a technique for analyzing various types of big data. The most commonly used technique in data mining is association rules. There are three types of interestingness measures used to generate association rules. One is objective type, the other is subjective type, and the other is semantic type. As with the confidence, if the value of the Hellinger measure is different for the two cases (x→y, y→x), it is difficult to decide which one to decide whether to create a association rule. In this paper, we proposed a symmetric Hellinger measure that can replace the elementary association thresholds and the Hellinger measure. In particular, we confirmed some useful properties by some example data. As a result, we found that a symmetric Hellinger measure was more suitable as a association threshold than them because it was larger than or similar to Hellinger measure.