Forecast Of Daily Rent Price Of Shared Accommodation Based On Hedonic Price Method

According to the "Shared Accommodation Development Report 2019" released by the National Information Center,the average annual growth rate of shared accommodation income from 2015 to 2018 is about 45.7%,which is 12.7 times the income of traditional accommodation industry.However,the growth rate has slowed down significantly compared to the 76% and 137% year-on-year growth rates in 2013 and 2014.The development of shared accommodation has entered the current critical "adjustment stage" from the original "high-surge" stage.This means that with the continuous increase in the number of shared accommodations in China,the degree of competition among landlords is increasing day by day.How to make reasonable and accurate price predictions in the fierce market environment has become the research problem of this paper.This paper takes Beijing Airbnb shared accommodation as the research object.(1)Using the method of literature review,starting from the characteristics of housing sources,homeowners,location,reputation,leasing,time characteristics to integrate 50 factors affecting the daily rental price of shared accommodation.(2)Based on the Hedonic Price Theory,on the basis of significantly different daily rental prices for working days and non-working days,the semi-logarithmic stepwise regression method and quantile regression method are used to build working day/non-working day shared accommodation hedonic price model so that we can obtain the main factors that affect the daily rent price of working days/non-working days and the specific impact of various factors on the working day/non-working day prices.(3)Based on these main factors,Random Forest and LightGBM are used to build working day/non-working day shared accommodation daily rent price prediction model so that the landlord can accurately predict the price of the house while knowing which factors are affecting the daily rent price.The results show that:(1)Daily rental prices of working days/non-working days are mainly affected by 30/33 factors,and the top10 factors of both are concentrated in the characteristics of housing supply,location and reputation.The difference is that non-working day prices are influenced by more factors,such as "whether there is a kitchen" and "whether the landlord is a superhost ".(2)Different factors have different effects on the price.Taking the number of accommodation as an example,for every additional person,the average price of working days/non-working days increases by 9.7% and 10.6%.And high-priced housing is more sensitive to this factor.(3)The average price of non-working days is 10.5% higher than that of working days,and the average daily rent price of holidays is 15.6% higher than that of weekends,which means that the landlord can adjust the rent price of the whole day in different time dimensions to obtain higher income.The novice landlord can follow this standard to make price adjustments.(4)The LightGBM Price Prediction model is better than the Random Forest Price Prediction model through the 10-fold cross-validation method.Among them,the average prediction accuracy of the LightGBM Prediction Model for the daily rental price of working day housing and non-working day housing is 71% and 82% respectively.