Research On Prediction Methods Of Commercial Real Estate Cost

In recent years,China's commercial real estate industry is developing rapidly,and many new types of commercial real estate are constantly emerging.With the continuous development of the industry,the construction cost estimation of commercial real estate is particularly important.Under the traditional cost estimation method,the cost prediction of commercial real estate is mostly based on a kind of quota issued by the government and combined with certain adjustment parameters.This calculation method is subjective and lack of persuasion,which can no longer meet the needs of industry management.Therefor it is urgent to introduce new methods and ideas to improve the efficiency and accuracy of cost prediction.Through the retrieval and analysis of docments in this field at home and abroad,it is found that over the years most research has focused on traditional residential projects and municipal projects,while the research on cost prediction for commercial real estate projects is relatively rare.Commercial real estates,especially office building projects have special characteristics such as few sample cases,unique design,complex system and difficult construction,so the cost prediction of such projects is very different from that of traditional residential or municipal projects.Therefore,the research on the cost prediction in this field has great theoretical significance and practical value.With the development of computer science and artificial intelligence theory,some artificial intelligence algorithms are gradually applied into the cost prediction field and a lot of research results have been achieved.For example,the research of BP neural network in the traditional housing and municipal projects is a hotspot in recent years.BP neural network is suitable for project cost prediction with large sample amount,and there may be a fitting phenomenon in the small sample prediction scenario,which affects the prediction accuracy and stability.that means BP network may not be the best choice for the commercial real estate,especially the office building projects.Based on the traditional BP neural network,this paper tries to introduce the generalized regression neural network algorithm(GRNN)into the field of commercial real estate cost prediction.Compared with BP neural network,GRNN network is more suitable for small sample amount prediction and has better performance in approximation,training speed,data classification and many other aspects.In addition,as there are many factors affecting the building cost,If all of the factors are used as the input data of the prediction model,the model will be too much complex.In this paper,principal component analysis(PCA)is introduced to reduce the dimension of the original sample data.PCA transforms the original data into a new data matrix,which can retains most of the original information,and new data components are independent to each other.Finally,in order to verify the prediction accuracy and stability of PCA-GRNN model,this paper also introduce the support vector machine method and BP network model,and use the same sample data to train and predict respectively.Through comparative analysis of relative error,average relative error rate and root mean square error and other indicators,it proves that compared with the commonly used cost prediction methods in recent years,PCA-GRNN method has higher accuracy,stability and easier operation in the scenario of commercial real estate cost prediction with small samples and multiple influencing factors.The model not only has the same ability of self-adaptive and self-learning with BP neural network,but also being suitable for small sample scenes and need less artificial parameters.Above research has several innovations as follows: firstly,the mathematical modeling method is introduced to predict commercial real estate cost;secondly,the small sample cost prediction model based on PCA mthod and GRNN network is established;thirdly,two classical prediction models(SVM and BP network)are introduced in the research process for comparison Research.The main conclusions are as follows: firstly,it is proved that it is feasible to introduce the mathematical method into the field of commercial real estate cost prediction,which means we have a new choice to solve the industry problems;secondly,a small sample prediction model based on PCA-GRNN is established,which provides a better theoretical exploration for similar small sample project cost prediction.In this paper,it is the first time to apply support vector machine method in the field of commercial real estate cost prediction.It is very hard to choose the right core function and set proper parameters,and these work really deserves further research.