基于随机森林的房地产项目风险评价

利用随机森林算法，基于房地产公司的视角，研究了 48 宗住宅地块的潜在投资风险，建立了包括 3 个维度 18 个指标的房地产项目风险评价指标体系。结合对J公司高管的访谈结果，进行了随机森林模型分析，发现该模型精确度高达95.6%，能够有效识别房地产项目的风险，得出了 5 个最重要的指标:项目区位、教育资源、城镇可支配收入、城镇化率、地区吸引力。并利用 10 折交叉验证得出了对于本研究数据，随机森林相对于 k 最近邻和支持向量机误判率更低的结论。

Risk Assessment in Real Estate Projects Based on Random Forests

Using random forests algorithm and based on the perspective of J Real Estate Corporation，this paper studies thepotential investment risks of 48 pieces of local residential land. A risk evaluation index system for real estate projects containing 3dimensions and 18 indicators was established. In this study，based on the results obtained from the interviews of executives of Jcorporation，an analysis using the random forests model is carried out. It is found that the model has an accuracy of 95.6%，which caneffectively identify the risks of real estate projects. The five most important indicators are also obtained, including project location，education resources，urban disposable income，urbanization rate，and regional attractiveness. Finally，using 10-fold cross-validation,this research concludes that the random forest has a lower misjudgment rate compared to the K nearest neighbor and the supportvector machine.