Harmaini, Fandi Husen and Mahmuddin, Massudi (2016) Empirical analysis of classifiers and feature selection techniques on mobile phone data activities. ARPN Journal of Engineering and Applied Sciences, 11 (10). pp. 6252-6258. ISSN 1819-6608

PDF Restricted to Registered users only Download (248kB) | Request a copy

Abstract

Mobile phones nowadays become ubiquitous device and not only a device to facilitate communication, with some addition feature of hardware and software.There are many activities can be captured using mobile phone with many of features.However, not all of these features could benefit to the in processing and analyzer.The large number of features, in some cases, gives less accuracy influence the result. In the same time, a large feature takes requires longer time to build model. This paper aims to analyze accuracy impact of selected feature selection techniques and classifiers that taken on mobile phone activity data and evaluate the method. Furthermore, with use feature selection and discussed emphasis on accuracy impact on classified data of respective classifier, usage of features can be determined. To find the suitable combination between the classifier and the feature selection sometime is crucial. A series of tests conducted in Weka on the accuracy on feature selection shows a consistency on the results although with different order of features.The result found that combination of K* algorithm and correlation feature selection is the best combination with high accuracy rate and in the same time produce less feature subset.

Item Type:	Article
Uncontrolled Keywords:	feature selection, classification, mobile cellphone activities, machine learning.
Subjects:	Q Science > QA Mathematics > QA75 Electronic computers. Computer science
Divisions:	School of Computing
Depositing User:	Dr. Massudi Mahmuddin
Date Deposited:	16 Jan 2017 07:05
Last Modified:	16 Jan 2017 07:05
URI:	https://repo.uum.edu.my/id/eprint/20602

Actions (login required)

View Item