An enhanced approach for LOF in data mining

Many techniques are available to find outliers. Out of those, local Outlier Factor (LOF) is quite efficient and well researched outliers mining algorithm. LOF quantifies, how much outlying an object is, in a given database. We proposed, in this paper, a modification in k-distance and named it m-distance that enhances the performance. k-distance is the distance between object and its k^th nearest neighbor, while m-distance is mean distance of an object and its k-distance neighborhood, increased by user supplied value λto increase performance. Modified algorithm is named as MLOF. The evaluation on real dataset shows that the proposed modification on LOF detects outliers more effectively.