Employee Turnover: A Novel Prediction Solution with Effective Feature Selection

This study proposed to address a new method that could select subsets more efficiently. In addition, the reasons why employers voluntarily turnover were also investigated in order to increase the classification accuracy and to help managers to prevent employers’ turnover. The mixed subset selection used in this study combined Taguchi method and Nearest Neighbor Classification Rules to select subset and analyze the factors to find the best predictor of employer turnover. All the samples used in this study were from industry A, in which the employers left their job during 1st of February, 2001 to 31st of December, 2007, compared with those incumbents. The results showed that through the mixed subset selection method, total 18 factors were found that are important to the employers. In addition, the accuracy of correct selection was 87.85% which was higher than before using this subset selection (80.93%). The new subset selection method addressed in this study does not only provide industries to understand the reasons of employers’ turnover, but also could be a long-term classification prediction for industries. Key-Words: Voluntary Turnover; Subset Selection; Taguchi Methods; Nearest Neighbor Classification Rules; Training pattern