Research On Location Promotion Algorithms Across Social Networks

In recent years,with the rapid development of online social networks and the widespread application of mobile positioning technology,location-based social networks(LBSNs)play an increasingly important role in information dissemination.In LBSNs,users can sign in at a location to record their visit history and share their feelings,thereby affecting their friends to also sign in at that location.Based on this feature of location-based social networks,many companies use it as the main battlefield for product promotion and marketing,i.e.,selecting seed users on location-based social networks and using their influence to promote services.Location promotion is defined as the problem of maxmizing influence in LBSNs,i.e.,how to find the best target user group in LBSNs,and use the information released by them to quickly promote the products and services to more users in LBSNs.It can help companies get the most economic benefits with the least advertising costs.In the location promotion research,how to model the diffusion process of information in the location social network and how to select seed users to make the promotion effect best are the two main research contents.However,most of the existing research on location promotion is only conducted in a single social network,ignoring the cross-network information diffusion ability of overlapping users(users who join multiple social networks at the same time),and lacks of mining user behavior patterns.For seed selection algorithms,most researchers use greedy algorithms or heuristic algorithms to select seed users.Compared with the heuristic algorithm,the seed set selected by the greedy algorithm can affect more users,but its time complexity will be higher.How to apply it to the multiple social networks is worthy of further study.To this end,this paper proposes an efficient location promotion method,which includes a cross-location social network propagation model(C-LBSN-PM)and a reverse reachable sketch seed selection algorithm with pruning(DC-IMM).The C-LBSN-PM dissemination model fully considers the influence of overlapping users in two different social networks and proposes a mechanism for blocking information dissemination between overlapping users according to the actual situation.When measuring the probability of information diffusion in location social networks,factors such as user historical visit interest,user check-in mobility,check-in similarity between users,and the attenuation of information intensity over time are comprehensively considered.The DC-IMM seed selection algorithm reduces the number of sketches that need to be generated on the basis of the low time complexity and high accuracy of the reverse reachable sketch algorithm,and reduces the space complexity of the seed selection algorithm.On the other hand,in the existing location promotion research,many researchers have neglected that it takes a certain time for information to spread in the location social network,and the structure of the location social network has changed during this time.In response to this problem,this paper proposes a dynamic location promotion method based on fusion network prediction.First,the problem of dynamic location promotion based on fusion network prediction is defined,and then a method that can predict the fusion social network is proposed.In this method,the anchor nodes in the fusion social network are split first,and then the multi-layer LSTM and autoencoder are used to predict the network structure after splitting.Finally,information dissemination and seed selection are carried out based on the predicted social network structure.Finally,this paper selects the location social network YELP and splits it into two social networks as the experimental data set.This paper first determines the value of the pruning distance in the reverse reachable sketch seed selection algorithm with pruning and the parameters of different characteristics in the information dissemination model of the cross-location social network.Subsequently,the C-LBSN-PM propagation model proposed in this paper was evaluated by three indicators of influence range,accuracy and recall rate.The experiment results show that the accuracy and recall rate of the propagation model proposed in this article are higher than the comparative propagation model selected in this article,while the propagation range is guaranteed.Subsequently,this paper evaluates the selection algorithm in terms of scope of influence,accuracy,coverage and time complexity.The experiment results show that the running time of the DC-IMM selection algorithm is shorter,and the accuracy of the selection is also higher.Further more,this paper verifies the dynamic location promotion method based on the fusion network prediction,selects seed users in the predicted fusion location social network,and compares the above evaluation indicators with the selection results of the real fusion network.The experimental result show that the dynamic location promotion algorithm proposed achieves good results.