Food and Restaurant Recommendation System Using Hybrid Filtering Mechanism

Abstract

The application of machine learning and Artificial Intelligence system in the food industry is not more profound than the other sectors. Those available systems do not fully answer the expectations of customers, lacks user personalization's. This paper presents the recommendation system for restaurants and food using a hybrid filtering mechanism. Multiple filtering mechanisms were applied on datasets to recommend food and restaurants for customers. The dataset was collected from recognized machine learning repositories of the U.S.A. government. Feature extraction and sampling are done on the datasets to test the performance of the system. This paper will try to answer user personalization preferences by applying Hybrid mechanisms. The recommendation was based on different customer preference like ratings, top sale, discount, weather condition etc. This paper combines content-based and collaborative based filtering mechanisms to provide the user with full functionalities of the recommender systems. It will adopt a hybrid system from the two mechanisms for effective implementation of the recommendation. For the development of this paper, data collection was done from known repositories, and feature extraction was done to filter out the unnecessary data. Sampling techniques were also applied to the dataset to distinguish between train and test data. 70% of the total dataset were allocated for training the system and 30% for test purpose. Sampling was also conducted in the model test stage by assigning 30% for evaluation purpose. To evaluate the performance of the proposed system, machine learning algorithms such as random forest, gradient boosting, decision tree, linear regression and K-Nearest neighbor were applied. The final performance of the model is so promising that it achieved an 83.5% success rate. Model loss and accuracy were also conducted, and the best fitting algorithms were selected. Based on the final result, the random forest algorithm shows significant performance with 0.859 accuracies and 0.1193 ...