SmartSifter (version 0.1.0)
Online Unsupervised Outlier Detection Using Finite Mixtures with
Discounting Learning Algorithms
Description
Addressing the problem of outlier detection from the viewpoint of statistical learning theory. This method is proposed by Yamanishi, K., Takeuchi, J., Williams, G. et al. (2004) . It learns the probabilistic model (using a finite mixture model) through an on-line unsupervised process. After each datum is input, a score will be given with a high one indicating a high possibility of being a statistical outlier.