基于机器学习的员工生产效率预测

Research on Productivity Prediction of Employees Based on Machine Learning

摘要: 本文从量化分析的角度出发，建立了多种机器学习模型对UCI数据库中的服装厂员工生产效率数据进行了研究。本文首先从物质激励、工作负荷、生产事故、目标促动四个维度构建了多维指标体系来对目标变量实际生产效率进行预测。为了获得良好的分类预测效果，本文建立了不同核函数配置的支持向量机、核密度朴素贝叶斯和随机森林共6个机器学习模型，其中随机森林的测试集分类正确率最高，为83.20%。其次，本文对初始随机森林模型进行了参数优化，优化随机森林模型的泛化能力有所提升。最后，本文根据随机森林特征重要性，分析得出了影响实际生产效率的最重要的因素。

Abstract: From the perspective of quantitative analysis, this paper established a variety of machine learning models to study the productivity data of garment factory employees in the UCI database. A mul-ti-dimensional index system from four dimensions of material incentive, workload, production ac-cident and target actuation was constructed to predict the actual productivity of target variable. Then six machine learning models including SVM with different kernels, kernel density naive Bayes, random forest were established to obtain satisfactory results. The test set’s classification accuracy of random forest was the highest, which was 83.20%. Then, the parameter of the initial random forest was optimized, and the generalization ability of the model was improved. Finally, based on the fea-tures importance obtained by the random forest model, this paper analyzed and concluded the most important factors affecting the actual productivity.