Interest point recommendation method based on heterogeneous attribute network representation learning

Abstract

The invention discloses an interest point recommendation method based on heterogeneous attribute network representation learning, and the method is suitable for recommending interest points to a useron a sign-in data set with abundant description information of the interest points and strong seriality. The method comprises the following steps of firstly, constructing a directed weighted heterogeneous attribute network based on a social network of the user and the sign-in data set; secondly, acquiring node attribute embedding information in the heterogeneous network based on text description of interest points, and acquiring meta-path embedding information of nodes based on random walk of multiple meta-paths in combination with a self-attention mechanism; then, fusing the attribute embedding of the node and the embedding information of the plurality of meta-paths, and performing representation vector learning of the node based on a heterogeneous skip_gram; and finally, based on the similarity of the representation vectors, performing accurate recommendation of the next interest point according to the time and place of the target user.

Description

technical field [0001] The invention relates to a network interest point recommendation method, in particular to an interest point recommendation method based on heterogeneous attribute network representation learning. Background technique [0002] In recent years, the widespread application of various smart mobile devices such as tablets and wristbands has significantly enhanced people's ability to generate and collect data. Every corner of life may spew out a large amount of data. How to use existing or It is very challenging to propose new techniques to mine valuable and interesting information from these massive data. Under the background that personalized recommendation is getting more and more attention from academia and industry, personalized recommendation technology in the field of location-based social network (LBSN) has been continuously enriched and developed. [0003] In the field of personalized recommendation based on LBSN, the recommendation of points of int...

AI Technical Summary

Problems solved by technology

[0003] In the field of personalized recommendation based on LBSN, the recommendation of points of interest is relatively complicated. There are many factors that need to be considered comprehensively when predicting the user's next check-in, such as time, current location, and descriptiveness of points of interest. Contextual information such as text and social relations. In addition, since the generation of check-in information is more expensive than online reviews, some users’ check-in records will be sparse, which will also increase the difficulty of recommendation.

At present, there are a lot of research work on point of interest recommendation. Zhang Yun (CN 107341261 A) et al. proposed a user collaborative filtering model based on spatio-temporal features, but it lacks effective mining of the check-in order of points of interest, and is not suitable for Deal with the sparse check-in records; Yu Dongjin (CN 109948066 A) et al. proposed a point-of-interest recommendation model based on heterogeneous information networks and meta-paths, which can effectively combine meta-paths and rich context information to mine user preferences, but Difficulty coping with cold start issues for POIs and users

[0004] Through the analysis and summary of some existing point-of-interest recommendation methods, the traditional methods have shortcomings in the following aspects: 1) Ignoring or not paying attention to the impact of each user's check-in sequence, resulting in inaccurate recommended points of interest and reducing user experience ; 2) Usually, the interest point representation is learned as a low-dimensional vector representation. In fact, each interest point plays two roles: one is the user's current position, that is, the starting point of interest; the other is relative to the previous one. POI, as the location where the user will check in, is the end POI; 3) Ignoring or not paying attention to the unique attributes of POI, such as geographical location, category and description information, etc., effectively using these peripheral information can further improve the accuracy of recommendation ; 4) It is difficult to deal with the cold start problem of users or points of interest

Images