Research On Forecasting Models Of Housing Rent Based On The Data Of Housing Rent Platform

With the development of economy and the flow of population between cities,the demand for housing in Beijing is increasing.In addition,the inclination of policies and the development of rental housing market make the rental housing economy more and more prosperous.The price of rental housing will affect the demand of the demand side.For the housing provider,it is particularly important to find the price position of its own housing source,so as to make the rent in a reasonable range of market price.The demand side can also know the actual value of potential housing sources through relevant forecasts.And there are also differences in price positioning between cities,so it is necessary to make differentiated pricing according to the city's own situation.In order to make rent pricing more reasonable and find out the influence of influencing factors on the pricing,it is necessary to build a prediction model to study and analyze the rent in Beijing.In particular,for the relevant housing rental platform,according to the different characteristics of the model on the impact of rent and its own positioning,it is more convenient to find out more accurate measurement factors and target price of housing in the future,and also more convenient for housing rent pricing.This paper divides the data from website into four categories:location characteristics,rental characteristics,community characteristics and housing characteristics.Making adjusting parameters strategy,and the Random Forest and B-P neural network prediction model are applied to the rent prediction of Beijing.Three indexes are used to compare the prediction accuracy with the traditional statistical models such as multiple regression model and Ridge regression model.This paper finds out the relationship between the importance and potential influence of some different influencing factors on housing rent,and uses the model with the best prediction effect for empirical application.The disadvantage of this paper is that the modeling data used is based on the data of a rental housing platform,but the data feature dimension can be expanded again,for example,medical surrounding,entertainment,education dimensions and seasonal factors.At the stage of modeling,when using software to build Random Forest model and B-P neural network model,the calculation speed of the model is slow,which needs a lot of calculation and time cost.In the empirical analysis,firstly,the domestic and foreign research of the housing rent prediction model is introduced.Secondly,the related theories of the housing rent prediction model are introduced,including the Random Forest model,multiple linear regression and Ridge regression model,the neural network model,the construction of prediction indicators and the Crawler technology.The data set studied in this paper is from a rental housing platform website.Firstly,the original data is preprocessed to generate data set,including the removal of irrelevant variables and handling with missing values and outliers.After the feature engineering correlation analysis,including the feature cleaning,normalization and virtual coding,the feature analysis is used to analyze the distribution of variables in the data set.The feature processing and construction are used to transform the data set into structured data that can be used by regression models such as Random Forest model and B-P neural network model.Finally,according to the correlation analysis,variance analysis and the importance ranking method of Random Forest variables,the important features are selected from the whole feature subset.The Random Forest model,the multiple linear regression model,the Ridge regression model and the B-P neural network model are selected to train the rent data,and the prediction effect is compared.The results show that the Random Forest model plays an integrate advantage,and the three indicators to measure the prediction effects are better than the other three prediction models,reflecting that the Random Forest model has a higher prediction accuracy in housing rent prediction.In addition,the use square and rental type of the house play the most important role in the rental pricing.Among them,when the use square reaches a certain degree,its effect on the rent pricing is weakened.In the case of renting with someone else,its effect on rent pricing is easily affected by other factors,while in the case of renting alone,its effect on rent pricing is stable and low.Secondly,the indoor living enviroment such as the number of bedrooms,the number of roommates,the location characteristics of the region and the distance to the subway station also have a strong impact on rent pricing.The overall facility environment of the area has the least impact on the rent pricing.So on the whole,the most significant influencing factors of Beijing housing rent are housing characteristics and rental characteristics,followed by location characteristics,and finally community characteristics.