Balanced Association Threshold of Dependency Factor

Among the knowledge discovery techniques used in data mining, association rules have received significant research attention. One of the general topics in association rule is development of good interestingness measures. Interestingness measures are the cornerstone of successful applications of association rule discovery and a meaningful scheme of interestingness measures may be based on user involvement. Among them, confidence is the most frequently used, but it has the drawback that it can not determine the direction of the association. The dependency factor expresses the degree of dependency, but considers only positive confidence. So, we proposed a new interestingness measure called balanced dependency factor considering simultaneously positive and inverse dependency factor, and compared the changing shape of three types of dependency. The results showed that if confidence is greater than a marginal probability, dependency factor is greater than 0. And if the sum of confidence and inverse confidence is less than 1, two item sets have negative association and the balanced dependency factor is less than 1. In contrast, if the sum is greater than 1, this measure is also greater than 1 and they have positive association.