Anomaly Detection Using Pagerank Algorithm

— Anomaly detection techniques are widely used in a various type of applications. We explored proximity graphs for anomaly detection and the Page Rank algorithm. We used a different PageRank algorithm at peak in proximity graph collection of data points indicated by vertices, gives results a score quantifying the extent to which each data point is anomalous. In this way we requires first forming a density calculating using the training data, it was high calculative intensive for sets of high-dimensional data. In the case of mild assumptions and appropriately chosen parameters, we explored that PageRank probability in point-wise consistent density imagines for the data points in an asymptotic sense and decreased computational effort. With that heavy betterments in case of executing time are experienced while maintaining similar detection performance. This way is computationally tractable and scales perfectly to huge high-dimensional data sets.