Design And Implementation Ofhouse Recommendation System Based On Knowledge Question And Answer

In recent years,with the development of the Internet and the progress of cities,online rental software is a necessary tool for many migrant workers to find accommodation in cities.However,with the development of the Internet and the progress of the city,the number of properties has gradually increased,making it difficult for people to find a property they are satisfied with.The recommendation system is a good solution to the problem of choosing from such a large number of resources.However,different recommendation algorithms have their own drawbacks,for example,collaborative filtering recommendation algorithms are more intelligent but face the problem of cold start for new users,and content-based recommendation algorithms need user’s historical data set as the basis.Therefore,a single recommendation algorithm cannot accurately describe the user’s preferences,and for interactive systems with a large data base,using collaborative filtering algorithms with a large computational system will largely affect the response speed of the system.It can be seen that the recommendation system faces three major challenges,namely: 1)low accuracy of single recommendation algorithm;2)large computational system of recommendation algorithm;3)lack of user preference features,which affects the accuracy of recommendation.To address the above problems,this thesis conducts research in two aspects: recommendation algorithm and user preference feature mining.First,in terms of recommendation algorithms,the research of price-based kmeans++,TF-IDF-based content recommendation algorithm,and ALSbased collaborative filtering recommendation algorithm is carried out.Based on the above three recommendation algorithms,the hybrid recommendation algorithm is implemented by combining the listings based on different price,orientation and other multiple feature dimensions to generate the listing information and by using the listing information after price clustering as the base listing data and calculating the similarity between the base listing data to generate a multi-model merging method of the listing similarity set.Secondly,in terms of user preference feature mining,a Chinese named entity recognition model with BiLSTM-CRF for listings is introduced.bi LSTM has good convergence and CRF layer ensures the legitimacy of the output prediction labels.The recognition accuracy of the model is improved by manual construction of question and answer templates and model building training,and after several experiments of parameter adjustment,we are able to achieve an entity recognition accuracy of 98% for the model in the field of rental properties in Nanjing.Finally,on the basis of the above research,a rental recommendation system based on knowledge Q&A is designed and implemented to verify the theoretical results of this thesis.In summary,the hybrid recommendation algorithm combining multiple recommendation algorithms and the research implementation of perfecting user preference features through knowledge Q&A in this thesis can effectively improve the recommendation accuracy of the rental system and enhance the user interaction experience with the continuous expansion of the data volume in the rental field and the increase of the rental demand,which has certain theoretical and application value.