An Effective Approach to Classify Fraud SMS Using Hybrid Machine Learning Models
Nidhi Agrawal, Abhishek Bajpai, Kumkum Dubey, Bdk Patro
Abstract
Due to the increasing proliferation rate of fraud messages, authors build a more accurate and precise model to detect it. This model is split into two stages: the first is the classification of SMS (Short Message Service) messages using a hybrid model, and the second is the examination of URLs. Our method majorly focuses on the classification phase here we design a hybrid model to improve the accuracy and precision of the classification model. Our hybrid model used the Naive Bayes Classifier, Random Forest, and Extra tree classifier. Individually, the accuracy and precision of Multinomial Naive Bayes (MNB) are 95.9% and 100%, Random Forest (RF) is 97.0% and 99.0%, and Extra Class Classifier(ETC) is 97.7% and 99.1% respectively. In the analysis of the dataset, our model got an accuracy of 96.86% and a precision of 99.366%. It is noticeable that our hybrid model outperformed machine learning approaches in terms of performance.