Abstract

Abstract- A dataset may contain objects that do not comply with the general behaviour or model of data .These data objects are outlier. Outlier detection has attracted increasing attention in machine learning, data mining and and statistics literature. A well-known definition of "outlier" is given as "an observation which deviates so much from other observations as to arouse suspicions that it was generated by a different mechanism," which gives the general idea of an outlier and motivates many anomaly detection methods Common general techniques for data classification include both unsupervised and supervised pattern classification methods. Some common approaches use clustering instead of simple feature selection, linear discriminant methods,neural networks and support vector machines Feature selection forms an important subset within the much larger area of data classification. Correctly identifying the relevant features in a data is of vital importance to the task of text classification. Our objective would be to actively select instances with higher probabilities to be informative in determining feature relevance so as to improve the performance of feature selection without increasing the number of sampled instances. Active sampling used in active feature selection chooses instances in two steps: first, it partitions the data according to some homogeneity criterion; and second, it randomly selects instances from these partitions.

Authors and Affiliations