No Detail Found!

MACHINE LEARNING FOR CONTENT-BASED DETECTION OF SPAM SMS IN LOCAL LANGUAGES: A PRELIMINARY CLASSIFICATION OF ROMANIZED URDU MESSAGES

Author(s)
M.S. Khan University of Engineering & Technology, Mardan, Pakistan.
S.I. Ayub University of Engineering & Technology, Mardan, Pakistan.
M.A. Khan University of Engineering & Technology, Mardan, Pakistan
M. Afaq Sarhad University of Science & IT,Peshawar,Pakistan
F. Tila Bacha Khan University,Charsadda,Pakistan
Abstract
Recently, a drastic growth is being observed in mobile device services and its usage. Majority of the people rely on their cell phones for all types of communication. This increase in the use of mobile devices has led the network providers to provide more services on low cost. Short Message Service (SMS) is a widely used service among these services. SMS provides a very affordable medium of communication and hence is used by all sorts of businesses and organizations to stay in touch with their clients and many use the service for advertisement purposes. As a result, the average user of a mobile phone is receiving an increased number of SMS messages not relevant to her. The conventional spam SMS detection and avoidance techniques work better for English language but for SMS typed in local languages, such techniques are not so efficient. This study considers content-based spam SMS detection for local languages. As a preliminary investigation, a dataset of actual SMS typed in Romanized Urdu was collected for a number of cellphone users. The dataset was pre-processed and Support Vector Machine (SVM) was trained using the dataset for classification of spam and normal messages. The preliminary results are promising as the content-based classification for Romanized Urdu messages using SVM gives an accuracy of 96.8%.
Publication Details
Page(s) 189-193
DOI DOI not available
Published Journal: Pakistan Journal of Science, Volume: 71, Issue: 4S, Year: 2019

Request the full text from PASTIC

Keywords
support vector machine (SVM), Spam, Machine Learning classifiers, Short Message Service (SMS)
0

Citations

2

Downloads

63

Views