ENHANCED SMS SPAM DETECTION USING BERNOULLI NAIVE BAYES WITH TF-IDF

Abstract

The use of mobile text messaging for communication is increasingly widespread, with Short Message Service (SMS) experiencing significant growth over the last decade. Consequently, the increase in SMS usage has led to a concerning rise in SMS spam, presenting substantial challenges for users and service providers. This study proposes a novel method for detecting SMS spam by combining Term Frequency-Inverse Document Frequency (TF-IDF) with Bernoulli Naïve Bayes (BNB) algorithm. The approach employs the use of TF-IDF for comprehensive feature extraction and the classification capabilities of the Bernoulli Naïve Bayes Algorithm. Through experimental validation employing TF-IDF for feature extraction and the BNB algorithm for classification, the results demonstrate high accuracy (98.36%), precision (99.19%), and a notable Matthews Correlation Coefficient (MCC) of 0.93, showcasing superior model performance compared to existing benchmarks.  Likewise, the proposed model shows efficient processing time (0.22 seconds). By combining strengths of TF-IDF and BNB, the approach offers effective SMS spam detection, surpassing the performance of traditional and deep learning classifiers. This research contributes valuable insights towards enhancing SMS security, thereby increasing trust between users and service providers.