Online Short-term Rental Price Forecast

Because of its low price,unique features and new interpersonal experience,online short rent is becoming more and more popular among travelers.In this thesis,the statistical model and arithmetic model of online short-rent house price are established to predict the online short-rent house price based on the online short-rent house source data of Tujia.It provides a method for landlord to evaluate the performance-price online,make better decisions.The specific work is summarized as follows:1.Collect and process data.Use Octopus software to crawl online short-term rental information in Chengdu,including dynamic price information and static information(house type,interior structure,service guarantee information,etc.).These data are cleaned(missing values,processing of outliers,normalized processing)and feature construction(continuous variable discretization,categorical variable dummy transformation)to make it effective for building model data.2.Descriptive statistical analysis of the data.Descriptive statistical analysis of the four aspects of internal short-term rental housing(internal structure,location factors,service assurance information,evaluation factors)master the distribution of indicators,and analyze the rules that affect the price fluctuation of online short-rent.3.Construct a linear regression model of online short-term rental prices.The traditional linear regression model was used to predict the online short-term rental price,and the model effect was evaluated.The optimized stepwise regression model was used as the base model to select the features.4.Construct an algorithmic model for online short-term rental prices.The regression tree,bagging,boosting,random forest and neural network model were used to predict online short-term rental price based on the feature selection.Finally,the NMSE value was used to evaluate the effect of each model.The results showed that random forest test set had the lowest NMSE value,reaching 0.2,the prediction effect is the best.This thesis aims to use the online short rental data to understand the fluctuation law affecting the price of online short-term rental housing,to further predict the price of online short-term rental housing,to make decision support for consumers,to provide reference value for merchant pricing,and to have important practical Application significance.