استفاده از داده کاوی هیبریدی و تحلیل خوشه ای یادگیری ماشین به منظور پیش بینی نرخ جایگزینی متخصصان حوزه فناوری

This study applies clustering analysis for data mining and machine learning to predict trends in technology professional turnover rates, including the hybrid artificial neural network and clustering analysis known as the self-organizing map (SOM). This hybrid clustering method was used to study the individual characteristics of turnover trend clusters. Using a transaction questionnaire, we studied the period of peak turnover, which occurs after the Chinese New Year, for individuals divided into various age groups. The turnover trend of technology professionals was examined in well-known Taiwanese companies. The results indicate that the high outstanding turnover trend circle was primarily caused by a lack of inner fidelity identification, leadership and management. Based on cross-verification, the clustering accuracy rate was 92.7%. This study addressed problems related to the rapid loss of key human resources and should help organizations learn how to enhance competitiveness and efficiency.

Improving the organization efficiency of human resources has been a key research topic for a long period of time. The purposes of this study were to discuss why Taiwan technology enterprises cannot retain talented employees and to find ways to prevent turnover of these employees and increase the competitiveness of the companies. Therefore, we use a discrete technique to discuss enterprise evolution. This study focused on employees between 20 and 39 years old who contribute to the high turnover in Taiwan. We examined turnover using a measurement scale based on real, local data. Reliability and validity tests were used to ensure data dependability. Other results can be revealed by applying computational intelligence clustering analysis techniques for data mining (i.e., self-organizing map (SOM) combined with back-propagation neural network (BPN)). SOM–BPN, which can be applied to reveal characteristics related to turnover trend clusters, is expected to offer reliable data to support the decision making of company policymakers. In an effort to determine the clustering accuracy hit rate in SOM–BPN model, the variables were integrated and selected. This paper contributes to the forecasting literature by developing valid and reliable variables based on information obtained from prior literature and experts in the field. Moreover, this study developed an approach based on clustering analysis, SOM and neural network clustering, to determine the accuracy hit rate.

Previous studies on turnover tended to explore the factors behind turnover, but they rarely predicted turnover inclinations. Therefore, they could not be used to immediately and effectively avoid the loss of key employees. This study solved this problem by using cluster analysis in data mining. According to the characteristics of the clusters, we found that employees value certain factors, such as the internalized identifications of supervisors, business support, leadership, management, growth and self-realization. In terms of turnover tendencies, the influence of supervisor commitment (or loyalty) was more significant than organizational commitment, which was distinct from other turnover research results on Western countries. Managers should not assume that the employees are less likely to leave when they actively fulfill their duties or when companies increase salary and benefits. In addition, when companies strictly follow procedure and rules and neglect the human side of the workplace, supervisors can accumulate loyal subordinates and then encourage all of them to leave for another firm. Clearly, it is much more important to cultivate loyal employees than abide by each and every rule. Due to time and cost considerations, this study still can improve forecasting efficiency and research application by use other methods. Thus, we propose that future scholars develop studies that do not consider turnover rates based on dichotomy. Thus, we might include fuzzy theory and multiple criteria decision-making to find additional influential factors with respect to turnover. To more effectively filter out the valid questionnaires, we might include traps and evaluate valid questionnaires through the use of a design program, which would save a significant amount of time.