Real Estate Project Risk Assessment Method Based on Modified Support Vector Machine

This paper presents a novel real estate project risk assessment method utilizing modified support vector machine. The process of real estate project risk assessment is divided into four steps: 1) beginning step 2) planning step, 3) Implementation step, and 4) final step, and there are 19 indexes in our proposed index systems. In this paper, we exploit the multi-SVM classifier to make risk assessment which can solve the limitation of a single SVM classifier. In the multi- SVM classifier, training data can be mapped to the high-dimensional space by the mapping function and slack variables. We divide the initial training set into several non-overlapping subsets, and a corresponding classifier is designed for each subset train. Given a sample vector, we utilize a single SVM classifier to obtain the category labels. Afterwards, the cluster membership of the given vector to a specific cluster is calculated. Hence, the finally classifying results can be obtained through the data fusion process using multiple SVM classifiers. To make performance evaluation, a series of experiments are conducted based the testing data collected from ten real estate projects. From the experimental results, we can see that the proposed algorithm can effectively assess risks in real estate project than other methods.