家政服务员的顾客满意度预测及影响特征分析
Prediction of Customer Satisfaction of Domestic Helpers and Analysis of Influence Characteristics

以数据和算法为基础对家政服务员的顾客满意度情况进行识别和预测，可以为提前甄别满意度欠佳的员工提供信息支撑。通过随机森林算法构建顾客满意度预测模型，并基于现实工作流程，将属性划分成不同类别，分析在服务前和服务后的不同阶段，预测顾客满意度的关键特征，运用SHAP模型分析变量对模型的影响方向及特征之间的交互关系，归纳出特征变量的重要程度。结果表明融合心理行为属性的家政服务员顾客满意度预测模型效果更佳，在家政服务员接单前，技能数量是预测顾客满意度最重要的特征；而在接单后的阶段，做单量、合作费、星级、准时率是最关键的特征。本研究结合了机器学习算法与SHAP模型，改善了模型性能，为现代服务行业的满意度提升提供理论支持。
The identification and prediction of customer satisfaction of domestic staff based on data and algorithm can provide information support for identifying employees with poor satisfaction in advance. The random forest algorithm was used to build a customer satisfaction prediction model, and the attributes were divided into different categories based on the real work flow. The key features of customer satisfaction were predicted at different stages before and after service. SHAP model was used to analyze the influence direction of variables on the model and the interaction relationship between features, and the importance of the feature variables was summarized. The results showed that the combination of psychological and behavioral attributes had a better predictive effect on the customer satisfaction of domestic staff. Before the domestic staff took orders, the number of skills was the most important feature to predict customer satisfaction. In the stage after receiving an or-der, doing a single quantity, cooperation fee, star rating, and punctuality rate are the most critical characteristics. This study combines machine learning algorithm and SHAP model to improve model performance and provide theoretical support for the improvement of satisfaction in modern service industry.