Sistema de Recomendacao Baseado em Localizacao para Ambiente Fechado

Abstract

The Recommendation Systems and Localization Services are two sectors that have been growing immensely in the last few years, enabling solutions using those technologies to become very relevant and promising nancial returns. In this context, this project aims to develop a recommendation system that uses the user localization in a physical store to create a list of recommended personalized products that are also close to him, improving the overall user experience and the probability of product sales conversion. The Collaborative Filtering recommendation algorithm, one of the most famous and relevant in the eld, was used as base and modi ed to use the user location approach. A prototype was developed to show the users journey in a mobile application and also an API in Python to create the recommendation of the ten highest scoring products. The prototypes result was able to show from the initial page until the recommendation list, and the result of the modi ed algorithm implementation showed that not only the generated recommendation was coherent in the physical store context, but also an increase in score of over 16% and a rise even bigger in the third quartile in the performed tests, and also a mitigation in the cold start problem from the classic algorithm.