Data Analysis And Mining Research Of Urban Houses Based On Real Estate Big Data

With the continuous advancement of new-type urbanization in China,the scale of urban houses has exploded.Affected by urban resource allocation and the imbalance of housing supply and demand,housing contradictions across the country,especially in large cities,have gradually arisen.Housing registration transactions,housing vacancies,and housing prices have become major issues related to people's livelihood.It has become a top priority to effectively improve the delicacy management level of urban houses.Delicacy management of urban houses needs data as support.Unified registration of real estate has been carried out for many years in China,which has achieved the data consolidation of human,land,and house.Although there are increasing achievements in real estate registration,the registration data resources have not been fully exploited.Therefore,there is an urgent need to carry out data analysis and mining research of urban houses based on real estate big data,so as to provide a strong scientific basis and auxiliary decision-making services for the national and local macro-control of urban real estate resources,comprehensive development and utilization,and sustainable development.In terms of the current technology and social needs of data analysis and mining research of urban real estate,the concept and content of urban real estate big data are explored.Based on the unified registration data of urban real estate,relevant data derived from the use of urban houses is linked and integrated to build a database of urban real estate big data.In terms of the unified registration and management of real estate and national macro-control needs,a statistical analysis is made on urban real estate big data from the perspectives of human,house and business.The model and algorithm of data mining of human-house relationship,housing vacancy rate,and housing price forecast are studied.With the data analysis and mining of urban real estate data in City X as an example,descriptive analysis,predictive analysis and normative analysis of urban real estate data are conducted.The main research work and conclusions are as follows:(1)The research on the connotation and data organization methods of urban real estate big data: Firstly,the content of urban real estate big data is explored.Secondly,based on Kettle technology,registration data of urban houses is extracted from the unified registration database of real estate.Linked data derived from the use of urban houses is integrated through the real estate unit number.Data is loaded into the HBase database with Sqoop technology.Therefore,the organization and storage of urban real estate big data can be realized,and a data base for the implementation of big data analysis and mining of urban real estate is established.(2)The research on the model and algorithm of data mining of urban houses: With urban houses as the research object,the classification of urban houses is studied and the content of data mining and analysis of urban houses is explored.Statistical analysis methods based on urban housing registration data are adopted to conduct a descriptive analysis of the status quo of urban houses and their registration business.And the method of calculating housing vacancy rate is proposed based on the rule classification.Dimension reduction analysis of the factors affecting urban housing price is made based on the principal component analysis.Besides,price forecasting models of second-hand houses are respectively built according to the Lasso algorithm,Random Forest Regressor algorithm,XGBoost algorithm,and Stacking algorithm.(3)The statistical analysis of urban housing registration data and case study of housing price forecast: With the data analysis and mining of urban real estate in City X as an example,a diversified analysis of the unified registration data of urban real estate in this city is made based on statistical analysis,including but not limited to analyses of proprietors,types of urban houses,transaction and mortgage status,registration business and so on.Dimension reduction is performed on the characteristics of urban housing price with PCA.Price forecasting models of second-hand houses are respectively built according to the Lasso algorithm,Random Forest Regressor algorithm,XGBoost algorithm,and Stacking algorithm.Also,forecast accuracy of the models are compared based on these five indicators.The empirical research results show that as a single model,the accuracy of the model based on Random Forest Regressor algorithm is significantly higher than the other two models.The model based on XGBoost algorithm is significantly better than the other two models in terms of MAE,RMSE,RMSLE and absolute error.The five indicators of the fusion model based on Stacking algorithm are all better than that of the single model,thus giving the model smallest forecast error and enabling it to predict the price of second-hand houses more accurately.