Recommender Systems for Restaurant Choice Based on Analysis of Photos from a Gallery

This paper is dedicated to the recommender system field in a restaurant business using neural networks to analyse users’ photos to improve people’s decision-making process of choosing restaurants according to their preferences. To achieve the goal of predicting user taste in restaurants based on his photos from the phone gallery the convolutional neural network, as well as hybrid system were used as a basis for analysing and classifying the data. For training and testing the algorithms the YELP dataset was the best choice. It provides photos, reviews and categories for local businesses, but some additional work to classify all types of the cuisines had to be done. The paper provides comparative analysis of different methods of cuisine classification, based on photos examination, and then presents a prototype of a software system for recommending restaurants based on the analysis of all photos from a phone gallery. Moreover, the work investigated the costs and profit of such an approach for business and the willingness of users to give their information to the system.