Design And Implementation Of Online House Selection System Based On Hybrid Recommendation

In recent years,with the continuous advancement of Internet technology and digital informatization transformation,the real estate industry,as one of the pillar industries of the national economy,is gradually evolving from upstream developers to midstream service platforms.Especially under the impact of the Covid-19,the industry chain accelerates the upgrade from "Offline service economy" to "Online experience economy".At the same time,with the rapid development of the national economy,the urbanization process is accelerating year by year,and the demand for housing is also steadily increasing.Therefore,combining Internet technology to create a high-level house buying service experience platform has become a trend in the industry,and it is also the best way to go.To this end,the online house selection recommendation system came into being.However,in view of the complexity of house feature information,the traditional recommendation model is particularly insufficient in both the processing capacity of data sets and the cross matching of multiple features.In order to solve the above problems and improve the users experience of online room selection.This thesis combines traditional recommendation algorithms and deep learning models to design and implement an online house selection system based on hybrid recommendation.Specifically,this project designs a house recommendation model based on Wide & Deep algorithm on the home page of the system.In the meanwhile,in order to improve the exposure rate of houses and click rate of users,on the detail page of house,a similar house recommendation model Content-based &Embedding is designed,and a guess you like recommendation model based on the Deep FM algorithm is also designed.In the offline model training experiment based on Tensor Flow framework,the Precision rate of the model based on Wide & Deep reached79%,the AUC reached 78%,the Precision rate of the Deep FM-based guess you like recommendation model reached 82%,and the AUC reached 85%.It shows that the recommendation model designed in this project has good classification ability and can recommend the houses for users.In addition,in order to efficiently implement the hybrid recommendation model designed in this thesis,the feature engineering of "blood" of the model is designed in detail based on the housing and user meta information,and the feature construction is more effectively completed based on the Spark computing engine,and then based on Tensor Flow Serving completes and deploys the online service of the model.Meanwhile,so as to improve the overall experience of users and the integrity of the system,the system uses the mainstream framework like Spring Boot to build the backend service,Vue framework builds the front-end service,and based on Redis to storage the online features for improving the access efficiency of the system.Furthermore,it also adds personal center module and housing information management module to improve the overall function of the system,to ensure the ease of use and maintainability of the system.This subject is based on the project during the internship of author.With the help of the tutor,the function development and testing of each module has been completed,and it is running normally,achieving the expected goals and effects of the system.