基于非线性组合模型的P2P网贷平台危机预警研究

针对P2P网贷平台现金流较大、利润率较低和财务数据获取困难的特点,构建基于平台交易真实数据的危机预警评价指标体系和组合预测模型.将传统的财务评价指标转换成网贷平台交易数据指标,运用邻域粗糙集属性约简的方法对采集的数据指标进行降噪和约减处理,再基于机器学习理论引入神经网络、支持向量机和Logit回归等模型对数据进行训练.通过分组进行单模型和组合模拟预测,提高了新的破产指标下各模型预测的准确率.

Crisis Prediction for P2P Platform Based on Nonlinear Combination Model

Based on the characteristics of asset-light, large cash flow, low profit and difficulty in financial data acquisition, evaluation index system and the combination prediction model for crisis prediction based on the real data of platform transaction were constructed. The traditional bankruptcy financial evaluation index was transformed into the net loan transaction data index and the collected data was reduced by the method of neighborhood rough set. On this basis, the data was trained with the machine learning theory which introduced neural network and support vector, logit regression machine model and so on. Finally, the data was divided into single group and combination group for predictive simulation and the accuracy of each model under the new bankruptcy index was obtained.