Opinion based intelligent recommender system

The study of sentiment analysis on social media posts can be used to analyse human emotions towards certain brands, topics, or products. Collaborative Filtering (CF) is a technique used to create personalized recommendations based on the preferences of other similar users. In this paper, we proposed a system design to incorporate sentiment analysis and CF to come up with a recommender system based on texts on social media. Sentiment analysis was performed using various forms of two machine learning models — Multinomial Naïve Bayes and Long short-term memory (LSTM) to learn users’ sentiments towards different products. User-user CF was then applied to estimate the ratings of unseen products for users based on other users with similar tastes. Products with top predicted scores were then be recommended to users. Our finding had shown that the proposed Multinomial Naïve Bayes model using Term Frequency – Inverse Document Frequency (TF-IDF) was most effective in classifying human sentiments towards a product. Our user-user CF method allowed more relevant product recommendations to users since it was derived based on users who had similar tastes like them. This project can be extrapolated to real-life applications, such as e- commerce recommender systems, by recommending products to users based on their social media content.Bachelor of Engineering (Computer Engineering