Research On The Real Estate Evaluation And Factors Influencing Based On Multimodal Weighted Mixture Model

In recent years,with the rapid development of Chinese economy,the real estate industry plays a vital role in it.The change of real estate value will also reflect the change of national economic development,and the fluctuation of national economy or the national macro-control of real estate will also affect the trend of real estate value.Therefore,in-depth research on real estate has great significance to national development and national life.However,in the real market,there are many kinds of factors that affect the real estate value,so it becomes very difficult to evaluate the real estate value.Based on the real estate data of Changchun,this paper conducts an in-depth study on the two issues of evaluating the real estate value and exploring the influencing factors of real estate value through multimodal feature extraction.Evaluating the real estate value and finding the factors influencing on the real estate value,on the one hand,can have a macro control over the whole real estate economy,and on the other hand,it has an important reference value for housing investors.In this paper,a multimode weighted mixture model is proposed to evaluate the real estate value and identify the influence of factors.This method can mitigate the adverse effects of real estate heterogeneity.Using the model proposed in this paper,we can evaluate the unit price of real estate according to the corresponding characteristics and explore the different influences of different factors on the real estate value.To be specific,we first extract the features of the three modes of housing attribute,space attribute,and time attribute from the real estate data through data processing method and machine learning algorithm,and obtain the corresponding representation vectors.Then,we use the principal component analysis algorithm to obtain the low-dimensional representation of the representational vectors.The low-dimensional representation reduces the computational cost of learning,avoids the potential catastrophe of singular matrix inversion in the process of estimating model parameters with the expectation maximization algorithm,alleviates the coupling among features as well.Next,we propose an adaptive weighted strategy,model the multimode representation vector after dimensionality reduction as a multimode weighted mixture model,use the adaptive weighted strategy to obtain the feature influence of each feature,and the feature influence was introduced into the estimation of the house price.We use the expectation maximization algorithm to solve the hidden variables in the model,after training the model,we obtain the evaluated house price and the ranking of the importance of factors affecting the real estate value according to the feature weights learned.Finally,in order to verify the effectiveness of the model,we conduct a large number of experiments on real-world real estate data sets in Changchun.The experimental results demonstrate the robustness and accuracy of the algorithm proposed in this paper.