Analysis Of The Neural Network Prediction Model Of Real Estate Prices In Xi’an

As an important indicator to measure the economic level of a region,real estate prices reflect on the one hand the relationship between supply and demand in the real estate market in the region,and on the other hand reflect the level of the real estate’s own characteristics and location characteristics.With the development of the social economy and the hot real estate market,the investment attributes of the real estate industry have gradually been enlarged,and the general rise in real estate prices has formed a contradiction with the residential needs of residents.The government has successively introduced corresponding policies to curb the negative impact of high housing prices,such as proposing the slogan "No real estate speculation",establishing a reference price mechanism for second-hand housing transactions,etc.It can be seen that real estate prices not only affect the transaction behavior of both parties in the real estate transaction.It also affects the government’s formulation of relevant policies on the real estate market.Therefore,the research on real estate prices,especially the prediction of housing prices,is of great significance.This article takes the transaction prices of second-hand houses in Xi’an as the research object,and uses "Lianjia.com" as the data source to use the Houyi collector to crawl the transaction data of second-hand houses in Xi’an.The area includes Beilin District,Jingkai District and Qujiang District in Xi’an.,High-tech Zone.Then 34 indicators that affect house prices were selected from the two dimensions of the property’s own characteristics and location characteristics,and 6,961 data were obtained after preprocessing the data with the help of Python.Before constructing the multiple regression prediction model,the location feature index was reduced,and the 4 common factors extracted after factor analysis were used as the real estate location feature index,combined with the principal component analysis results to conclude: living needs and supporting facilities for transportation education are measured An important factor of location level.After that,five housing price prediction models were constructed using Python language: Ridge regression,Lasso regression,elastic regression,Xgboost regression tree,and BP neural network.It is found that the BP neural network prediction model has a mean square error of 0.0014 on the training set after 100 trainings,and a mean square error on the validation set of 0.0158,which is the smallest error in the above prediction model,which indicates that the BP neural network has a problem in housing price prediction.Has better performance.This study constructed a BP neural network prediction model for real estate prices in Xi’an,and used the results of factor analysis to give an evaluation method for the location feature level of real estate,which provided basis and method support for real estate price evaluation and prediction.The research results can not only give advice and guidance to both parties in real estate transactions,but also enable the government to formulate macro-control policies in a targeted manner to ensure the balance of supply and demand in the real estate market and healthy development.