Comparison Of Classification Algorithm In User Sentiment Analysis Of Getcontact Application In Online Fraud Prevention COMPARISON OF CLASSIFICATION ALGORITHM IN USER SENTIMENT ANALYSIS OF GETCONTACT APPLICATION IN ONLINE FRAUD PREVENTION

Abstract

Online fraud refers to various fraudulent acts carried out over the internet with the aim of fraudulently obtaining financial gain or personal information. We need to continue to spread awareness about the importance of security for ourselves and the people we know, where currently there are many different modes of online fraud. One application that is well known to the public is the GetContact application, which is an application designed to provide information about incoming calls, identify spam or fraudulent calls, and provide services related to a list of telephone contacts that have been registered by fellow users of the application. In this research, researchers will analyze the sentiment of comments from users of the Getcontact application by comparing the test results of classification algorithms, namely Naïve Bayes Classifier and SVM. This research process will begin with data sampling using the scrapping technique on Google Playstore and processing data from users of the Getcontact application using RapidMiner. After the preprocessing process and model testing with two textmining methods using algorithms, namely SVM and Naive Bayes, the evaluation and validation results show that Naïve Bayes has a higher level of accuracy than SVM. For Naïve Bayes, the accuracy value reached 82.97% with an AUC value of 0.500, while for SVM, the accuracy value was 78.00% with an AUC value of 0.926. These results show that Naïve Bayes is superior in classifying user comments on the Getcontact application on Google Play as positive and negative comments.