RSVRs based on Feature Extraction: A Novel Method for Prediction of Construction Projects’ Costs

The construction industry is an important basis for the development of China’s economy. The accurate prediction of the cost of the project determines the profitability of the project and the decision-making of the construction. Therefore, this paper combines Random Forest with Support Vector Regression, and puts forward a method of construction cost prediction based on this method which we call it Random Support Vector Regressions (RSVRs). In this paper, 22 kinds of features which are related to cost are extracted, then we use Kernel Principal Component Analysis to fuse features which are less relevant to the cost. Then 11-dimensional features are put into RSVRs for modeling. From the theoretical point of view, RSVRs gives a novel work flow that is fusing features with low correlation firstly and then performing regression. In order to avoid the defects of the regression method, the two methods are fused, so that each method can maximize its advantages and make up for the disadvantages of the other method. From a practical point of view, RSVRs can be used to predict the project cost, which not only improves the accuracy, but also immediately obtains the predicted result. As long as the appropriate training data is selected, the method can predict the cost of any area and any building.