Improving the Restaurant Recommendation Performance using Sentiment Analysis

Recommendation systems are effective information filtering systems that filter relevant information from a large amount of dynamically generated data and then provide prioritized information to the customers based on their preferences, likes and other factors. In this study, machine learning based restaurant recommendation system is developed and its performance is improved utilizing the reviews given by the customers. The Zomato Bangalore dataset used in this study is downloaded from Kaggle. Initially, reviews of various restaurants are given as input to classifiers namely Multinomial Naive Bayes (MNB) and K - Nearest Neighbor (KNN) to build models that can classify a given review as positive or negative. Then, similarity measures like Pearson correlation coefficient and cosine similarity are used to identify the reviews that are similar. Finally, the top ranked restaurants are then recommended to the user. Experiments were conducted to evaluate the performance of machine learning algorithms and similarity metrics. From the results obtained, it is found that MNB classifier with cosine similarity is able to recommend restaurants with higher accuracy.