GODAC: Graph-cut based outlier detection using ant colony optimization algorithm

Outlier detection plays an important role in data mining as outliers may contain some useful information in many applications. In this paper we propose a method of graph-cut based outlier detection using ant colony optimization algorithm. Both the correlation and the discreteness of the attributes are used to weight the data´s characteristics. We use the ant colony optimization algorithm to find optimal paths that will be component of a graph, and in this process we take both the distance and the distribution of the data into consideration which can contribute to more accurate results. On this basis we give the criterion of the outlier identification after we cut on the graph obtained according to the cutting criterion. Experiment results show that GODAC has good precisions in outlier detection.