A Deep Neural Networks model for Restaurant Recommendation systems in Thailand

Abstract

In the age of flooded information, Recommender Systems play a crucial role as long as consumers consume more content and submit more data. Many businesses have implemented Recommender Systems to assist users find items based on their previous interactions. Deep neural networks have demonstrated promising results in a variety of disciplines, including recommendation systems in the past few years. However, such studies ignore auxiliary information input. In this work, we purpose a deep recommendation system with neural networks which consists of deep collaborative filtering to learn user and item interaction latent factor and enrich the performance with textual information by using multi-layer perceptrons and combining these two models under our framework, called DNNRecs. Apart from our model framework, we also contribute a feature engineering method to create new features from review text by using technique tf-idf. Extensive experiments on one real-life dataset in Thailand demonstrate the effectiveness of the proposed model.