基于概率图模型的危险驾驶罪刑期预测

针对司法实践中对于可解释性及预测性能的需求, 本文提出了一种基于概率图模型的量刑智能辅助方法. 该方法以量刑要素为基石建立含有隐节点的概率图模型, 由极大似然准则估计刑期分布的参数, 进而计算分布的数学期望得到预测值. 关于危险驾驶罪的实验结果表明, 概率图模型的预测准确率优于基于决策树和神经网络等的模型, 且具有良好的可解释性.

Prison term prediction of dangerous driving based on probabilistic graphical model

To satisfy the actual demand for interpretability and prediction accuracy in judicial practice, we in this paper propose an intelligent sentencing method based on the probabilistic graphical model (PGM). This model is built on the cornerstone of sentencing factors. The parameters are estimated by using the maximum likelihood criterion, and the predicted values are obtained by calculating the mathematical expectation of distribution. Experimental result on dangerous driving shows that the prediction accuracy of the method is better than that based on comparison models, such as decision tree and neural network. Meanwhile, this method has good interpretability as well.