An Effective and Efficient Approach to Detect and Predict Outliers Visually
一种有效的可视化孤立点发现与预测新途径

Outlier detection is an integral part of data mining and is critical important to some areas such as monitoring of criminal activities in electronic commerce,credit card fraud,etc. Due to the topological structure and probabilistic distribution preserving nature,SOM (Self-Organizing Maps)has been used as a tool for mapping high-dimensional data into a two dimensional feature map and gaining some idea of the structure of the data by observing the map. Based on the analysis of the existing distance-based outlier detection algorithms,a SOM based approach to detect and predict outliers is proposed,which has an obvious superiority in scalability,predictability,interactiveness,conciseness. Experimental results on real database show that the SOM based outlier detection and prediction is effective.