This article is cited in 4 papers

Probability theory and mathematical statistics

Feature selection based on statistical estimation of mutual information

A. A. Kozhevin

Lomonosov Moscow State University, GSP-1, Leninskie Gory, Moscow, Moscow, 119991, Russia

Abstract: An algorithm to identify significant factors is proposed in the mixed model framework. It employs statistical estimation of mutual information. Consistency of this procedure is established. Numerical experiments demonstrating its accuracy supplement theoretical results.

Keywords: feature selection, mixed model, mutual information, conditional Shannon entropy, logistic regression.

UDC: 519.2

MSC: 62G20, 62H12

Received March 31, 2021, published June 23, 2021

Language: English

DOI: 10.33048/semi.2021.18.052

Bibliographic databases:

•