Urban Second-hand House Price Prediction Based On Big Data Technology

In the past two years,while emphasizing the acceleration of the urbanization process,China has continuously issued various control policies for real estate,and the future trend of the real estate market has gradually become the focus of everyone's attention.The acceleration of the current urbanization process is driving the development of the real estate market.In some popular cities,the real estate industry has even gradually developed into an important pillar industry.At this time,in-depth research on the real estate industry is of practical significance.Through in-depth understanding of the industry,general developers will study the value of urban undeveloped land and the positioning of built houses.This article is a more intuitive study of the price of second-hand houses in cities in China from the perspective of ordinary consumers.The results of the research can provide consumers with an important and reasonable reference basis when buying a house.The analysis and prediction of house prices have always existed in the real estate market,but the actual methods used so far are still limited to traditional methods such as statistical analysis or multiple linear regression.These methods have not reached the most in the current environment of data blowout and complicated social factors.Good results,and the processing effects of ordinary machine learning models can not meet the requirements of high-level institutions.At present,there are few cases where big data and deep learning technology are used in house price prediction.In view of the disconnection between this cutting-edge technology and practice,this paper combines the big data platform and uses deep learning methods for modeling.The combination of the technology and the application of housing prices,a hot issue of people's livelihood,is more reasonable to grasp the internal laws of the data,and further improve the accuracy of current house price forecasting methods.This article is to study the price of second-hand houses,but it has changed from previous research methods.The data used in this article is the real estate transaction data of China 's cities(Hefei,Anhui Province).All the house information and surrounding supporting information are from the real house transaction website and the official platform of the relevant department,and on this basis,a complete set of The system of influencing factors of house prices in order to ensure the practicality of the final model.The core experiment of this paper is based on the Big Data Spark platform,using time series-based LSTM and GRU models combined with time-series real estate data to predict housing prices.In order to prove the effectiveness of the method,multiple linear regression of machine learning models,decision trees,random forests,ARMA and deep learning models LSTM and GRU were selected,and these six models were used on platforms based on common platforms and big data Spark respectively.Carry out house price prediction experiments,compare the experimental effects of machine learning and deep learning on the same platform,and verify the impact of the platform on the experiment through the same model experiment results of different platforms.According to the experimental results,whether it is based on the ordinary platform or the big data Spark platform,the deep learning model has lower MAE and RMSE values than the machine learning model experiment,that is,the error is smaller.Compared with ordinary platforms,the MAE and RMSE index values of deep learning experiments based on the Big Data Spark platform are smaller and the accuracy is improved.This is mainly because deep learning is better at processing large data sets.On the one hand,automated cross-validation provides a better fit for the model.The time-consuming of all model experiments on the Spark platform based on big data is reduced compared to ordinary platforms.This is due to the distributed data storage of big data and the memory-based calculation method of the Spark framework.The experimental results of both LSTM and GRU are similar,but in terms of experimental time,the 23.53 seconds of the GRU model is less than the 26.43 seconds of the LSTM model,mainly because the GRU has a simplified loop structure compared to the LSTM model.In the end,it can be concluded that for time-series property data,the deep learning models LSTM and GRU are more accurate than other machine learning models.The LSTM and GRU models based on the Big Data Spark platform have more accurate prediction results.The Spark platform based on big data can increase the experiment speed and run more efficiently than ordinary platforms.The GRU model runs faster than the LSTM model.In summary,the GRU model based on the Big Data Spark platform can be applied to the real estate price prediction field,with higher prediction accuracy and better processing efficiency under the support of a large amount of data.