Research On Real Estate Price Prediction Based On Internet Search Data

Today,with the accelerating pace of life,the Internet has brought great convenience to people.Before making consumption decisions,Internet retrieval has become an important channel for consumers to obtain relevant information.Therefore,online search data can be used as a quantitative indicator of consumer attention,providing scientific and reasonable data for many research issues.As the pillar industry of China's national economy,the price of real estate is a key issue that people pay attention to,and it has also become a current research hotspot.However,the current research on real estate price prediction is flawed by the problems of research data and prediction models.Therefore,this paper takes Xi'an as an example and uses network search data to predict real estate prices.This article first analyzes the main influencing factors of real estate prices based on the equilibrium price theory,and reveals the relationship between real estate prices and online search data from a qualitative perspective.In the quantitative analysis of the correlation between real estate prices and online search data,this article uses subjective word selection and other methods to select 6 core network search data on the Baidu index website,and then builds a gray correlation model to calculate the relationship between the two.Gray correlation.Then,based on the core network search data,the initial database is built.In this paper,the secondary search and other methods are used to expand 119 network search data,and Spearman correlation analysis and time difference correlation analysis are used to filter out the first ones with high correlation with real estate prices.Network search data,and then use the principal component analysis method to synthesize the filtered network search data to eliminate collinearity between the data.Finally,using the integrated network search data indicators as variables to predict real estate prices,a long-term and short-term memory model,a support vector model,and a random forest model are used to predict real estate prices,respectively,and empirical tests are conducted using Xi'an data as an example.However,because each model has certain disadvantages,and the prediction results deviate from the actual real estate price,so this paper establishes a fusion model based on the establishment of three single models,and uses the gradient to improve the decision tree model to improve the prediction accuracy.The research shows that:(1)There is a strong correlation between real estate prices and online search data;(2)When using long-short-term memory models,support vector regression models and random forest models to predict real estate prices,the prediction effect of long-short-term memory models The best,followed by the random forest model,and finally the support vector regression model;(3)Use the gradient boosting decision tree model to predict that the fitting degree after fusion is closest to the actual real estate price,and the goodness of fit reaches 0.996.