Cross-Attention Mechanism Recommendation Algorithm Based on LSTM

Abstract

With the development of economy, tourism is favored by more and more people, and location-based point-of-interest recommendation (POI) begins to emerge. This paper proposes the lstm-based recommendation algorithm for cross-attention mechanism (L-Attention). Embedding method was used to accurately learn the location sequence and location information of users, and feature information of location data was extracted through LSTM network. At the same time, the cross-attention mechanism is introduced to conduct dynamic modeling of personalized time check-in sequence, which can integrate the user's behavioral preference, location sequence information and the influence of sequential location on next-poi, so as to improve the accuracy of recommendation. This article uses Foursquare as an experimental data set. The results show that compared with the traditional recommendation algorithm which only considers the popularity of POI and the deep learning recommendation algorithm which only considers the timing, the accuracy of the proposed algorithm is greatly improved.