Board game hybrid recommendation method based on multi-source heterogeneous data

Along with more and more board games and more and more board game users and more and more serious data sparseness problems, a traditional matrix decomposition method only uses scoring data, and the performance of the scoring data is limited when a scoring matrix is sparse, so the invention provides a board game hybrid recommendation algorithm based on multi-source heterogeneous data; multi-sourceheterogeneous auxiliary data is fused on the basis of scoring data, so a problem of insufficient expression of game features and user features due to sparsity of a scoring matrix is solved. Improvements are provided on the basis of a traditional probability matrix factorization (PMF) algorithm, and the improvement points include 1, enhancing representation of game feature vectors by using game description texts and game attribute information, and 2, enhancing representation of user features by using features of games scored by users in combination with an attention mechanism. Experiments showthat compared with a basel