A new approach for feature transformation to Euclidean space useful in the analysis of multispectral data

摘要

Presents a simple and an efficient procedure to transform the n-d data to m-d data where m can range from 1 to n-1, based on feature standard deviation (FSD), feature range value (FRV), feature column grouping (FCG) and dynamic feature sorting (DFS). This procedure requires the computation of standard deviation of features. The reduction is stopped at a value of m which gives maximum discrimination. This improves space and time requirements. The clustering tendency index (CTI) is used to quantify the suitability of the transformation method. The efficacy of the algorithm is established by experimental studies made on various data sets.