An Experimental Study to Improve Due‐date Performance

Abstract

Due‐date performance (DDP) is a very important performance indicator for the companies. Thus, companies with a high hit rate would have greater competitive advantage; on the contrary, companies that delay customers' orders frequently would lose sales opportunities and reputations. Therefore, there were many academic studies and practical efforts to improve DDP in the past, but the problem of low hit rate still exists. In order to increase the hit rate, some companies have focused on reducing the variation, while others focus on production management, but is the real problem affecting the low rate variability or production management? This is indeed difficult to be validated through practice. Therefore, this study designed three scenarios, tested each scenario for 30 times, each test involved seven subjects. The tests were to provide counter‐evidence in the Job Shop environment without variation. If the variation is the main factor of affecting hit rate, the rate at this time should be good; otherwise, the assumption that variation is the main cause is rebutted. The results demonstrated that production management planning is the main cause, and the method of enhancing the hit rate is obtained during the test.

Optimal Decision Tree for Cycle Time Prediction and Allowance Determination

“…The last line of research is to maximize the hit rate of a policy. Wang et al [38] considered production management planning under the theory of constraint framework to improve the hit rate and compared the performance of three experimental scenarios of a job shop. Chen and Wang [12] framed DDM as a supervised machine learning problem, that is, a neural network was trained and predicted on training and testing datasets, respectively, and demonstrated its effectiveness over seven other policies in terms of hit rate.…”

“…Instead, a convex l-1 norm penalty on , namely Lasso, is added and solved [18] [33]. Xu et al [38] proved that Lasso regression is equivalent to a robust regression problem with a feature-wise uncoupled ℓ 2 -norm uncertainty set.…”