Recommendation Technology Summarize on Multi-agent Game Confrontation

Abstract

Game confrontation is an important research direction of multi-agent system, which will form massive strategic data. How to excavate and discover useful strategies to make game confrontation more efficient becomes a meaningful research direction. Firstly, this article introduces the mainstream recommendation algorithms, which are demographic-based recommendation algorithm, content-based recommendation algorithms, collaborative filtering based recommendation algorithm and combinatorial recommendation algorithm. Secondly, the implementation process of the recommendation algorithm is described, and then the most commonly used similarity calculation method for finding nearest neighbors is introduced. Finally, the difficulties and solutions in the current recommendation algorithms are introduced, such as sparsity of matrix, algorithm scalability, new user cold start problem and so on. Future recommendation research direction on data diversification and further improvement of recommendation algorithm.