Youtube spam detection framework using naïve bayes and logistic regression

Abstract

YouTube has become a popular social media among the users. Due to YouTube popularity, it became a platform for spammer to distribute spam through the comments on YouTube. This has become a concern because spam can lead to phishing attack which the target can be any user that click any malicious link. Spam has its own features that can be analyzed and detected by classification. Hence, enhancement features are proposed to detect YouTube spam. In order to conduct the experiments, a YouTube Spam detection framework that consists of five (5) phases such as data collection, pre-processing, features selection and extraction, classification and detection were developed. This paper, proposed the YouTube detection framework, examined and validate each of the phases by using two types of data mining tool. The features are constructed from analysis by using data collected from YouTube Spam dataset by using Naïve Bayes and Logistic Regression and tested in two different data mining tools which is Weka and Rapid Miner. From the analysis, thirteen (13) features that had been tested on Weka and RapidMiner shows high accuracy, hence is being used throughout the experiment in this research. Result of Naïve Bayes and Logistic Regression run in Weka is slightly higher than RapidMiner. In addition, result of Naïve Bayes is higher than Logistic Regression with 87.21% and 85.29% respectively in Weka. While in RapidMiner there is slightly different of accuracy between Naïve Bayes and Logistic Regression 80.41% and 80.88%. But, precision of Naïve Bayes is higher than Logistic Regression.