Personalized recommendation algorithm of books based on the diffusion of reader's interest

Abstract

The ever-growing books help readers acquire knowledge faster than ever before. But the huge scale of these resources also easily makes people fall into the dilemma of "Information-Explosion", which prevents the reader from easily locating the books that are really suitable for them. To alleviate this dilemma, we analyzes the principle of the "Networks-Based-Inference" algorithm (NBI), which is a classical heuristic algorithm for recommendation. We also proposes an improved algorithm—NBI algorithm using Interest Diffusion (NBI-ID), that derives from this classical algorithm. This improved algorithm inherits the advantages of NBI method in simplicity and effectiveness, and optimizes the allocation of initial information in the process of information diffusion with an interest related indicator. Thus increasing the efficiency of the recommendation results. Experiments on the GoodBooks dataset show that the proposed algorithm improves in accuracy, recall and diversity compared to the classic NBI, CF (Collaborative Filtering) and GRM (Global Ranking Method) algorithms.