Detection of Hoax Spread in The Whatsapp Group with Lexicon Based and Naive Bayes Classification

Abstract

Spreading hoax through WhatsApp social media can lead to different beliefs and can cause disputes for those affected. This paper proposes a hybrid model for finding hoaxes in the WhatsApp group using a combination of knowledge-based and machine learning approaches. This Hybrid model combines two methods namely Lexicon based and Naive Bayes Classifier which will be applied to the WhatsApp monitoring application. This research focuses on two main aspects namely word weighting using the lexicon based method and data classification using the Naive Bayes Classifier and Decision tree-j48 methods. The dataset used is conversation data that is crossed from the WhatsApp group. Based on the experiments that have been carried out, it is obtained the results of classification using Naive Bayes classifier of 86.670% data conversation not indicated hoaxes and 13.330% indicated hoaxes. The average value of the percentage of truth obtained more than 75%. The average value of the classification performance evaluation results in a precision value of 0.771, a recall value of 0.754, an F-measure value of 0.773