现有环境治理成本预测模型构建主要基于专家经验进行输入 输出指标的选取. 然而专家经验指标选取存在一定的主观性, 难以有效且 客观地区分关键指标信息对成本预测结果的影响. 同时, 考虑到现有环境治理 成本预测模型的复杂度以及可解释性问题, 文章将扩展置信规则库模型(EBRB)用 于环境治理成本预测, 并基于维度约简介绍三类不同的特征选择方法对EBRB进行 结构学习, 与EBRB模型的参数学习方法结合, 构建基于不同联合学习方法的EBRB环境 治理成本预测模型, 最后基于我国31个省份的环境治理数据, 分析不同联合学习 模式下的区域成本预测差异, 不仅有效降低预测模型的复杂度, 且能弥补现有基于 主观经验进行成本预测指标选取的不足.

The existing environmental governance cost prediction model is mainly based on expert experience to select input and output indicators. However, there has a certain subjectivity of indicator selection by expert experience, so it is difficult to distinguish the impact of key indicator information on the cost prediction results effectively and objectively. At the same time, considering the complexity and interpretability of the environmental management cost prediction model construction, this paper introduces the extended belief rule-based (EBRB) model for cost prediction and the dimension reduction by the three feature selection methods are applied to the structure learning of EBRB model, which combines with the parameter learning methods of the EBRB model to build a joint learning EBRB model for environmental management cost prediction. Based on the environmental management data of 31 provinces in the mainland of China, this paper analyzes the regional cost prediction differences under different joint learning modes, which can not only effectively reduce the complexity of the prediction model, but also make up for the shortcomings of the existing cost prediction by subjective indicator selection.