An Improved LOF Outlier Detection Algorithm

LOF (Local Outliers Factor) algorithm is a very classic anomaly detection algorithm. In order to detect the outliers more accurately, avoid too much testing error, and ensure the detection can be implemented relatively accurately in the data set without professional knowledge, on the basis of traditional LOF algorithm, an improved detection algorithm LOF Outliers is proposed. According to the different distribution densities of logarithmic data points, all the data point sets A1 that are most likely to become outliers are found out. Then, the information entropy weighted LOF algorithm is used to detect the data set to get the result A2. The point set A1 is intersected with A2 to get the final point set A which is the final result. The experimental results show that the algorithm is feasible, and it is more accurate and contains fewer false detection points.