【摘要】 针对消费品召回措施有效性、召回后的残余风险量化缺乏明确的评判标准等问题，提出基于随机森林的消费品召回效果评估方法。首先，运用CIRO层级评估模型从背景、行为、产出和影响4个层面分析召回效果的影响因子。针对召回活动产生的直接和增益效果，构建消费品召回效果评估指标体系。然后，根据召回效果优度值及残余风险量级分析，划分召回效果等级。引入随机森林算法确定指标的权重系数，构建消费品召回效果评估模型。最后，通过匹配度和相对误差验证，随机森林运用在该模型时的拟合预测值中，93.33%的样本在合理的误差范围内，说明随机森林算法适合并适用在消费品召回效果评估领域，该算法可以有效、快速地发现多个指标数据之间的潜在规律，进行准确预测。更多还原

【Abstract】 Facing with a lack of clear evaluation criteria for the residual risk quantification and effective consumer product recall measures after recall, this study proposes a random forest-based consumer product recall effect evaluation method. Particularly, the CIRO hierarchical evaluation method is applied to analyzes the factors of effect of consumer product recalls from the four levels of background, behavior, output and influence. Aiming at the direct and gain effects of recall activities, a consumer product recall effect evaluation index system is constructed. Then, according to the merit value of recall effect and the analysis of the magnitude of residual risk, the recall effect is divided. The random forest algorithm is introduced to determine the index weight, and the consumer product recall effect evaluation model is established. Finally, through the verification of matching degree and relative error, 93.33% of the samples of the fitting prediction value of the random forest used in the model are within a reasonable error range, which indicates that the random forest algorithm is suitable and applicable to the field of consumer product recall effect evaluation. The algorithm can effectively, quickly discover the underlying laws between multiple indicator data and make accurate predictions.更多还原