Urban POI Knowledge Discovery And Recommendation Optimization Based On LBSNs

With the rapid development of the mobile Internet,Location Based Social Networks(LBSNs)have penetrated all aspects of our lives.Tt is very important to discover valuable knowledge from LBSNs' data and to recommend Point of Interest(POI)to users in the face of a large amount of information.However,Different POIs contain different information because of their different visibility.How to extract POI knowledge and realize information inference between POI? In addition,different platforms contain different features of POI,how to realize the feature fusion of the same POI in different platforms? In addition,in the existing POI recommended research,it is very rare to consider the time accessibility of POI,and what time the POI is suitable for play is very important for the user experience,so how to make POI recommendations based on temporal feature? In view of the above problems,this study has done the following aspects:Aiming at the first question how to extract POI's knowledge characteristics from LBSNs and conduct POI information inference,We proposed a method of knowledge discovery based on LBSNs and extracted POI knowledge characteristics through natural language processing.Based on the POI knowledge feature,we proposed a POI similarity algorithm to construct the POI relation network.By experimental verification,our solution is a good way to extract POI semantic characteristics,POI network can realize the information transmission between the POIs and unknown information reasoning,through this step,we can have a better understanding of the POI which has less infomation.To solve the second question how to extract POI temporal feature from LBSNs and make time prediction for unknown POI,We first propose a POI temporal teature method,and then we put forward a joint learning algorithm cross LBSNs,fusing the multidimensional information and reasoning unknown knowledge(POI knowledge features and temporal feature),through the matrix decomposition we can study the temporal feature of unknown POI and POI knowledge featuers,Through experimental verification,we can extract the POI temporal feature in real time,and the feature joint learning scheme can well fuse the different characteristics of the POI,predict the unknown POI time distribution,and learn the time distribution of POI knowledge characteristics.To solve the third question how to make POI recommendations based on temporal features,On the basis of the time distribution of POI knowledge acquired by the above joint learning algorithm,POI recommendation strategy based on temporal feature was established based on POI knowledge characteristics.We verified the proposed algorithm on the real data set,and the experimental results show that our method has better performance in POI time prediction and recommendation.The contribution of this paper lies in the following three aspects: first,through the POI feature extraction,POI relation network construction method is proposed,which can not only keep the characteristics of semantic information but can better transfer information between POI and reason unknown POI knowledge;Secondly,this paper proposes a multi-dimensional information fusion and reasoning method based on the feature joint learning method of LBSN.Through matrix decomposition,the temporal feature of unknown POI and POI knowledge features,and the method can be extended to the fusion of multi-dimensional information and knowledge inference in other fields.Thirdly,a POI recommendation algorithm based on the temporal feature is established.This paper proposes a new travel content and mode of innovation,we should not only focus on the user's preference for POI content,but also consider the POI temporal feature.