Recomendação de produtos baseada em perfis de contatos de clientes e popularidade de produtos

Abstract

Customer Relationship Management (CRM) systems have a large amount of contact information. Knowing which products should be offered to these people is not a simple task, due to the large quantity and diversity among the products, so making personalized recommendations for each contact profile can help in the sales process. Based on this, this work presents a product recommendation system for contacts, using recommendation based on product popularity and recommendation using collaborative filtering. In the popularity-based recommendation, a ranking is created with the best-selling products, and the products at the top of that ranking are recommended. In the recommendation for collaborative filtering, two metrics are used: the Levenshtein distance and the cosine similarity. With each metric contact lists with similar profiles are created for each contact, and from these lists the products are recommended. Clustering is also applied to the data, using the cosine and Levenshtein metrics. At the end, the results of the recommendation methods are compared, where the methods that use clustering obtained better results.