A feature selection method based on Lorentzian metric

Yerzhan Kerimbekov, Hasan Şakir Bılge

2016

Closed access Metadata (CSL-JSON)
DOI 10.1109/bigdata.2016.7840791
Data confidence 70 medium
How this score is calculated

It is built from checkable properties of the record itself: whether the article has a DOI, how many independent sources confirmed its data, whether the required fields are filled in, and how recent the latest confirmation is. Disagreement between sources and a missing identifier lower it. A value of 80 or above means the record was published without further checking; 60 to 79 means it is published at reduced confidence, with a manual check advisable. A record verified by an editor by hand is never pushed below 90 by a recalculation. This scores our data about the article, not the article itself.

Abstract

Nowadays multidimensional data as a part of Big Data are collected in every organization and feature selection is one of the main approaches in terms of processing them with machine learning methods. In this study, firstly, a Feature Selection based on Lorentzian Metric (FSLM) is developed. The proposed method unlike from matrix multiplication in Euclidean space uses the Lorentzian analogue to calculate within and between class scatter matrices. These scatter matrices are functioned in the novel FSLM method as discriminative criterion. In calculation of the scatter matrix, covariance matrix in Lorentzian space should be properly defined. The validity and correctness of the novel FSLM method was tested over GESTURE, LSVT, MADELON, RELAX and SONAR data sets. The obtained results indicate that the proposed method has a significant improvement in terms of feature selection and classification success.