FLDS: Fast Outlier Detection Based on Local Density Score

The problem of outlier detection is one of the issues attracting much attention from researchers in the field of data mining and knowledge discovery from data (KDD). Detecting outliers in a data set is to find different points than the majority of the remaining points. Currently, many studies have been introduced to address this problem. One of the widely used algorithms is based on the evaluation of local density of data, for example LDS or LOF algorithm, however, the drawback of these methods is that the high computational complexity cost O (n2). In this paper, we propose a method for outlier detection based on k-nearest neighbors with the complexity of the method is O (n1.5), named FLDS. The basic idea of this method is to use a K-Means algorithm to devide dataset into k clusters, then find the singularity in this clusters. Experimental results show that the new algorithm is capable of detecting outliers similar to LOF, but the calculation time is faster about 20 times than LOF.