Research On Urban Residential Land Price Assessment Based On Artificial Intelligence Reasoning Model

With the continuous advancement of urbanization in China,the urban land economy accounts for more and more importance in the national economy.The core of urban land economy is urban residential land prices,so it is of great practical significance to accurately and timely assess urban residential land prices.The evaluation methods of traditional urban residential land prices have the disadvantages of long period,low efficiency and large errors.The residential land price data often has defects such as imbalance and small amount of data.Therefore,it is necessary to propose a rapid assessment method suitable for the characteristics of residential land price data.This paper proposes a land price feature extraction method based on migration learning,aiming at the shortage of total residential land price data and unbalanced distribution of categories,and then uses different land price assessment models for accuracy assessment,and finally explores models and methods for land price assessment.The main research work of this paper is as follows:(1)In view of the inadequacy of traditional land price feature selection and the subjectivity of feature quantification,this paper starts with business,transportation,education,infrastructure,environment,and other comprehensive factors to construct a land price impact factor system with wide coverage.Combined with the characteristics of each impact factor and its relationship with residential land prices,a quantitative system of residential land price factors in Shenzhen is established.(2)This paper proposes a feature extraction algorithm to solve the small amount of training data.By introducing the transfer learning technology and considering the correlation between the property price characteristics and the land price characteristics,the housing price feature transfer is applied to the extraction of residential land price characteristics,which makes the original land price data small and the feature extraction difficult to be solved,so as to realize the high precision assessment of the residential land price.(3)In the process of extracting land price characteristics using principal component analysis,this paper changed the selection method of principal components with a eigenvalue greater than 1 and tested the different principal components using cross-validation to find the main components suitable for residential land price classification..(4)According to the migration learning feature extraction method proposed in this thesis,we first use price data and deep belief network(DBN)to perform feature extractor training,then based on house price DBN model for land price feature extraction,and finally select three commonly used land price classifications.The model(Support vector machine,BP neural network,Random forest)analyzes the classification accuracy of the extracted land price features,and verifies the effectiveness of the residential land price feature extraction algorithm based on migration learning.At the same time,it is concluded that the random forest classification model is more suitable for the urban residential land price classification evaluation.This paper improves the land price feature extraction method.Experimental results show that the land price feature extraction algorithm based on migration learning has better classification accuracy than the commonly used principal component dimension reduction extraction method and linear normalization method.Compared with the principal component analysis(PCA)method,the proposed feature extraction method has an average accuracy of 10.9% and 4.73% higher than that of the principal component analysis method,and the optimal feature set classification accuracy has reached 90.28%.Therefore,the residential land price evaluation algorithm proposed in this paper can meet the actual land price evaluation accuracy.