Research And Implementation Of Collaborative Filtering Based Second-hand House Recommendation And Prediction

With the continuous development of Internet big data technology,people are mining data to find valuable information,which brings us more convenience but also brings many problems.Take buying a second-hand house as an example,thousands of listing data will bring trouble to home buyers.The recommendation system is an effective protection measure to recommend the items users want in time when facing the huge data information.In order to make good use of these data,this paper designs and implements a recommendation and prediction system for second-hand properties based on collaborative filtering using recommendation system technology combined with a second-hand property data platform to address the problem of too much information on traditional second-hand property websites.The main work of this paper and the innovation points are as follows:(1)This paper proposes a hybrid collaborative filtering method TACF(time and interests collaborative filtering)that combines time weights and user interest changes.The challenge of recommendation systems nowadays is the increasing number of users and items,which leads to the problem of low accuracy and sparse data.In this paper,we propose an improved hybrid collaborative filtering recommendation algorithm.Firstly,we establish the user interest distribution matrix to calculate the interest similarity among users;secondly,we introduce the time weight function to calculate the user rating similarity;finally,we combine the two similarity methods and use the improved prediction rating formula to calculate the similarity.This better reflects the changes of users’ interests,which can significantly improve the recommendation accuracy compared with the traditional recommendation algorithm and improve the recommendation quality in the case of sparse data.(2)For the traditional second-hand house price prediction problem,it is still difficult for the current market to effectively and accurately predict the house price problem.This paper combines a machine learning algorithm model to compare the traditional house price prediction problem on the derived with better predictability and accuracy.In some cases where there is less data on second-hand houses to compensate for the bias caused by sparse data,five different algorithmic models are predicted using cross-validation techniques to obtain five mean squared errors and standardized mean squared errors respectively,and the model with the best prediction results is selected.In order to avoid the shortage of prediction and evaluation methods to further ensure that the predicted second-hand house prices are more reasonable,the characteristic prices are also introduced to combine with the best prediction model algorithm to construct a price evaluation model,which ensures the accuracy and reliability of the final results of second-hand house prices.(3)Finally,the second-hand house recommendation system is designed and implemented based on improving the accuracy of the recommendation algorithm and second-hand house price prediction.The second-hand house recommendation system not only satisfies the users in terms of demand and functional experience,but also effectively improves the problems of low accuracy and sparse data in the recommendation system,helps users to quickly understand the desired property information,greatly improves the troubles when purchasing a house,and has certain practical application value.