The Research And Implementation Of Location Recommendation System Based On Sequence,Social And Geographical Factor

With the rapid development of information science and technology,recommendation systems are an important means to resolve information redundancy in the era of big data.The rise of mobile Internet in recent years has made location-based services more important.There are some problems and challenges in the field of interest recommendation that need to be solved urgently,and the research content of this article is based on this.The main tasks include the following:(1)Aiming at the problem of data sparseness commonly found in the point of interest recommendation problem,this paper proposes a model based on directed probability graph embedded representation learning.First use the original check-in data to construct a directed weighted graph,and then implement the alias sampling algorithm to complete sampling by probability distribution to achieve the effect of data enhancement;use the unique geographic attributes of the points of interest to correct the transition probability between points of interest,So that the trained vector can not only learn the sequence association between contexts,but also the geographic association between contextual interest points;finally,use the idea of transfer learning to pre-train on a large number of corpora to obtain embedded representations of interest points.(2)This paper proposes a deep neural network model based on attention encoder to learn the user's personal interest preferences.Thanks to the introduction of a recurrent neural network structure,the model can efficiently extract the user's interest distribution when processing check-in sequence data of different lengths;and the addition of the attention mechanism enables the model to adaptively interest users Change for dynamic learning.(3)In order to improve the accuracy of recommendation results and the personalized experience of users,this paper proposes a recommendation model based on Bayesian ranking.A sampling method based on geographic stratification is proposed.In addition,in order to accelerate the model's convergence speed,this model also proposes a small batch gradient descent method with geographic level perception.Finally,by optimizing the Bayesian ranking index,the recommendation list provided to users has better ranking quality.(4)Combining the above research content,this article designs and implements a complete point of interest recommendation system,which integrates factors such as context sequence factors,geographical factors,and interest preferences,effectively improving the accuracy and diversity of recommendation results.