Performance Evaluation of Linkage Pattern Mining Method

Linkage pattern mining is a data mining technique for discovering sets of patterns that appear repeatedly across multiple sequential data. We previously proposed a linkage pattern mining method using closed itemset mining and showed that it can effectively exclude pseudo patterns derived from noise. In addition, it was suggested that the extraction accuracy of this method was strongly affected by the correctness of frequent pattern extraction from each sequential data. The frequent pattern extraction process requires two parameters, a window width w and a minimum number of occurrences θ. However, our previous study has not performed comprehensive evaluation in various combinations of these parameters. In this study, we conducted a grid search for the parameter values that exhibit stable and high extraction accuracy. As a result, it is shown that both parameters should be set to smaller values.