Frequent Pattern Outlier Detection without Exhaustive Mining

International audienceOutlier detection consists in detecting anomalous observations from data. During the past decade, outlier detection methods were proposed using the concept of frequent patterns. Basically such methods require to mine all frequent patterns for computing the outlier factor of each transaction. This approach remains too expensive despite recent progress in pattern mining field. In this paper, we provide exact and approximate methods for calculating the frequent pattern outlier factor (FPOF) without extracting any pattern or by extracting a small sample. We propose an algorithm that returns the exact FPOF of each transaction without mining any pattern. Surprisingly, it works in polynomial time on the size of the dataset. We also present an approximate method where the end-user controls the maximum error on the estimated FPOF. An experimental study shows the interest of both methods for very large datasets where exhaustive mining fails to provide the exact solution. The accuracy of our approximate method outperforms the baseline approach for a same budget