A new outlier detection method for high dimensional fuzzy databases based on LOF

Abstract

Despite the importance of fuzzy data and existence of many powerful methods for determining crisp outliers, there are few approaches for identifying outliers in fuzzy database. In this regard, the present article introduces a new method for discovering outliers among a set of multidimensional data. In order to provide a complete fuzzy strategy, first we extend the density-based local outlier factor method (LOF), which is successfully applied for identifying multidimensional crisp outliers. Next, by using the left and right scoring defuzzyfied method, a fuzzy data outlier degree is determined. Finally, the efficiency of the method in outlier detection is shown by numerical examples.