SURVEY OF DISCRIMINATIVE FEATURE SELECTION AND CORRELATION MEASURE FOR HIGH DIMENSIONAL DATA

Omkar Bhalke, Lalit Kumar, V Patil
Journal of Innovative Research and Solutions   unpublished
The term "Feature Selection" refers to the identification of the subset, which contains the most useful features. This subset produces an original complete set of features as the compatible result. For reducing dimensionality, deleting irrelevant data, increasing results unambiguousness and improving accuracy of learning, the Feature subset selection method can be used very effectively. The use of cluster based Fast algorithm and the Fuzzy logic will improve the process. The FAST algorithm is
more » ... FAST algorithm is used for identification and removal of the irrelevant data sets. The processes that used in this algorithm, implements two different steps: First, Graph Theoretic clustering method, and second, Selection of Representative feature cluster. The features found in various clusters are comparatively sovereign. The probability of production of a useful subset and independent features is very high in FAST strategy based on clustering. The Minimum-spanning tree clustering method can be used for ensuring the efficiency of the FAST algorithm. The experimental studies can be used to further evaluation for the efficiency and effectiveness of FAST algorithm.
fatcat:y4hcekqkhndbjn2k3euqhoko4u