Research And Implement Of Rental Housing Recommendation Based On Graph Learning And Epidemic Perception

With the rapid development of the sharing economy,people have rerecognized the concept of shared rental housing,However,with the development of the Internet,the number and types of rental housing related information have increased dramatically,and the content it covers has expanded from ordinary physical information.to abstract information such as housing evaluation.On the one hand,this phenomenon enriches the tenant’s choice of rental housing,but on the other hand,it increases the user’s very serious information load.Recommendation systems are now widely used to mitigate information overload.However,at present,there is a lack of personalized rental housing recommendation software in the domestic market,and most of the existing products require users to manually query and screen the housing information,On the other hand,how to recommend rental housing for tenants under the background of the normalization of the epidemic,that is,to avoid the health risks caused by the location of rental housing in high-risk areas with the spread of infectious diseases as much as possible,also brings great challenges to research and has practical application value.In order to achieve the effect of recommending low-risk,high-reliability rental housing for tenants.It is necessary to make personalized rental housing recommendations based on tenant preference characteristics and the associated characteristics of rental listings,and predict the development trend of the epidemic based on realtime epidemic information and filter the recommendation results.Therefore,this paper mainly studies the rental housing recommendation algorithm based on graph representation learning and epidemic perception,and the main contents are as follows:(1)A cross-domain recommendation algorithm based on personalized migration of user preferences is proposed to solve the cold start problem.The algorithm constructs a feature encoder to find the user’s features in the source domain,then uses the meta-network to generate different bridge functions for different users,and finally initializes it for the target domain to complete the recommendation.Experiments based on the Amazon dataset show that the cross-domain recommendation performance of the algorithm is better than that of each benchmark algorithm,which proves the effectiveness of the personalized bridge function in cross-domain recommendation.(2)A rental housing recommendation algorithm that integrates geographic information and time series information is proposed,and the topological geographic impact is captured by establishing a geographic information map and using the messaging neural network.At the same time,graph neural networks are used to capture user preferences and explore sequence substructures in interactive timing graphs.Finally,the consistent learning of geographic and temporal information is carried out to achieve the ability of joint optimization of recommendation effect.Experiments based on the Airbnb rental order dataset show that the recommendation accuracy of the algorithm is higher than that of each benchmark,and the performance of the algorithm is adjusted to the best state through hyperparameter experiments,which proves the effectiveness of the new algorithm in rental house recommendation.(3)A cross-domain recommendation algorithm based on joint distribution is proposed to make cross-domain recommendation for warm startup users who have a large number of check-in data and rental house interaction data,and use federated distribution to capture the correlation of samples in different fields,so as to learn the overlapping features and specific domain characteristics in cross-domain recommendation,and use it for item recommendation in the two fields.Experiments on Amazon datasets show that the recommendation accuracy of the algorithm is better than that of the relevant benchmark algorithm.(4)An epidemic prediction algorithm based on spatiotemporal graph neural network is proposed,which models the graph structure for the new crown epidemic,and sets two different edges to fuse temporal attributes and spatial attributes to construct spatiotemporal infographics.The algorithm contains a two-layer graph convolutional network,and uses the hop connection model to connect the layers to avoid node information sparseness.Comparative trials on the NYT new crown case dataset and the Google aggregated mobile survey dataset show that the error of the algorithm in predicting the number of cases and new cases in each region is lower than that of the benchmark algorithm,which proves the effectiveness of the algorithm in the epidemic prediction scenario.(5)The sharing economy market is huge,but some information lacks effective integration,and the domestic market lacks personalized rental housing recommendation software that can cope with the normalization of the epidemic.Therefore,this paper designs and implements a personalized rental housing recommendation system with epidemic awareness function,and uses the algorithm proposed in this paper in this system to provide personalized recommendations for tenants and avoid the risk problems caused by the epidemic.