Track: Operations Management

Abstract

Major studies on cell production have placed an emphasis on technical factors such as machine order/layout, family part grouping, work-flow sequence, etc. However, it is still insufficient to investigate how human factors affect the productivity of production cells. Although many companies have introduced cell production systems and realized that productivity of cells varies greatly with workers, there is no effective measure to access workers’ aptitude assembly tasks in production cells. In order to assess the impact of workers’ aptitude on productivity, we have made a series of experimental studies and clarified that there are about 6% of workers who have very low aptitude and can be considered not suitable to assembly tasks in cell production. Meanwhile, intending to find the worker’s aptitude closely relating to her/his productivity, we designed a laboratory experiment, and administered 11 paper-and-pencil tests of the Japanese General Aptitude Test Battery (GATB) to the workers to measure their seven work-related aptitudes. After examining the partial correlations between the workers’ GATB scores and their productivity, we have clarified that the GATB scores can’t directly measure a worker’s aptitude towards the assembly tasks in production cells, but we could apply principal component analysis to obtain a component score that correlates significantly with the workers’ performance in cell production. However, this component score still can't be used to measure workers’ aptitude perfectly, in particular it can’t be used to identify the workers with very low aptitude.

This paper intends to propose an effective classification model to identify the workers with very low aptitude based on workers’ GATB scores. We pay special attention to make several new contributions to cell production research through the following examinations:

 (1) Based on workers’ GATB scores, we apply linear discriminant analysis and decision tree to construct several classification models and demonstrate that these models can’t identify the workers with very low aptitude with acceptable accurate.

(2) We apply random forest method to construct a classification model and examine its effectiveness, as well as the best choice of the parameters.

(3) We investigate the relationship between the detailed aptitude scores or standard aptitude scores of GATB and the workers’ classification, and clarify the aptitude scores that have a heaver impact on the classification.