Summary: | 碩士 === 國立宜蘭大學 === 多媒體網路通訊數位學習碩士在職專班 === 106 === In this knowledge-driven economic era, “people” has been regarded as an enterprise’s important asset. Voluntary employee turnover can cause inevitable cost and expense for the enterprise. This research aims to find the ideal model which can predict the rate of employee turnover and the reason behind it by means of data mining technology, thus enabling the head of department to comfort and retain the employees by using the reward system and communications so as to avoid the disadvantages brought by vital staff’s turnover.
This research collects company H’s 12 years of personnel files and divides them into Training data set to generate the prediction of employee turnover and then compare them with the actual employee turnover in the Testing data set, in order to estimate the accuracy of the predict model. And this research establishes 4 kinds of CART category models based on columns. Model Ⅰ is the basic identification data (29 columns), Model Ⅱ is the common column data of the company’s administrative document (12 columns), Model Ⅲ is the combination of the Model Ⅰ and seniority (30 columns), Model Ⅳis the combination of the Model Ⅱ and seniority (13 columns),
Model Ⅰ, the prediction accuracy of 96.14% is the highest; In Model Ⅲ, the prediction accuracy decreased slightly to 94.76%. Model Ⅱ had the lowest prediction accuracy, which was only 69.78%. Model Ⅳ, the prediction accuracy was improved to 82%.
Model Ⅰ analyzes the 14 reasons for employee turnover. Seven affecting reasons appear in 8 columns of the department organization category indicating the close relationship between employee turnover and the department’s organizational structure, leadership styles, factory culture and so on; the input of 2 columns in human resource category has no effects; 2 affecting reasons emerge in the six columns of personnel administration category; there are 5 affecting reasons in the 12 columns of personnel payment, among which “socialSecurityAllowance” is the first, “basedSalary” is the second. There are 9 columns contained by all the four Models, and 3 of them are the reasons affecting employee turnover: “originalFactoryID”, “currentFactoryID”, “basedSalary”. The former two are the information of the department the employee affiliates to, while “gradeID” exists in three Models and it refers to the employee seniority and job rating results.
|