Performance evaluation of feature selection algorithms on human activity classification

TULUM G., ARTUĞ N. T., Bölat B.

2013 IEEE International Symposium on Innovations in Intelligent Systems and Applications, IEEE INISTA 2013, Albena, Bulgaria, 19 - 21 June 2013, (Full Text) identifier

  • Publication Type: Conference Paper / Full Text
  • Doi Number: 10.1109/inista.2013.6577634
  • City: Albena
  • Country: Bulgaria
  • Keywords: Feature selection, human activity detection, ReliefF, t-score
  • İstanbul Yeni Yüzyıl University Affiliated: Yes

Abstract

In this work, four human activities were classified by using multi layer perceptron and k-nearest neighbours algorithm. Due to mass amount of data, two different feature selection methods, which are ReliefF and t-score, were applied to the data. The best result is obtained as 97.6% with 51 features selected by ReliefF. © 2013 IEEE.