A new outlier detection base on the self-organizing map algorithm

This article applies the method of artificial neural networks in the outlier detection, and gives one kind of new outlier detection method. According to the thoughts of the GHSOM algorithm and the GHTSOM algorithm, we make the improvement to the SOM algorithm. The improvement algorithm can be applied in the outlier detection, and this dissertation gives the different outlier detection examples and analyze the algorithm performance and the expansion ability, the performance is quite stable and adaptation is quite strong to the different date. Compares with the outlier detection use support vector machines, this method doesn´t need to choose the kernel function and adjust the parameter unceasingly, and it has very good adaptation to the change of the data distribution.