Análise e seleção de algoritmos de filtragem de informação para solução do problema cold-start item

Abstract

It is currently difficult for people to manage the growing volume of information available on the Internet. One example occurs when a consumer must choose among a lot of products available in e-commerce website. To offer their products and help consumers to choose the most relevant to them e-commerce companies use recommender systems. Such systems generally make use of techniques of information filtering to generate relevant recommendations. But when there are none or few data (ratings, purchases, searches, etc.) about items, the filtering algorithms do not provide good results. This problem is called Cold- Start Item. This work makes a detailed study of the most relevant information filtering algorithms of the last decade that can solve the Cold-Start Item problem and presents a analysis of these algorithms in order to identify the most promising one. This work is being developed in partnership with the company Chaordic Systems of Florianópolis and the results presented here will be used to develop algorithms for an recommender system developed by the company.