Sentiment Analysis and Fake News on Social Media

Soh, Chee Wei (2022) Sentiment Analysis and Fake News on Social Media. Final Year Project (Bachelor), Tunku Abdul Rahman University College.

Text SOH CHEE WEI_Fulltext.pdf Restricted to Registered users only Download (14MB)

Abstract

In this modern era, fake news has become a common thing that people will always do by using their phone or laptop to create and share some online false information or fake news to everyone on social media platforms. Fake news was simply designed to alter people’s interpretations and reactions to real news to impair their ability to distinguish between what is true and what is false. Therefore, fake news detection needs to be developed to solve the fake news issue. Apify is used in this paper to collect data which includes Facebook posts and engagements from Facebook about cosmetics products with specific keywords while following the CRISP-DM stages for the dataset process. There are 4 experiments with different features from fake news that will be conducted in this paper. Experiment 1 is using K-means with News Content-based and Social Context-based. Experiment 2 has the same features as experiment 1 but with One Rule and Skope-Rules Rule-based classifier. Experiment 3 is using K-means, One Rule and Skope-Rules classifiers with only News Content-based features and experiment 4 is using the same machine learning model as experiment 3 but with only Social Context-based features. There are only two Rule-based classifiers that are split into 2 different ratios of training and testing sets which are 70% training and 30% testing and 80% training and 20% testing. K-means will conduct with 2, 4, and 6 clusters. After the modeling, the 4 metrics of evaluation will be used in this paper which include accuracy, precision, recall, and F1-score to find out the best model with experiments. In short, One Rule is the best overall model in fake news detection with every experiment compared to the Skope-Rules classifier and K-means clustering