Data Mining Approaches for the Methods to Minimize Total Tardiness in Parallel Machine Scheduling Problem

This study examines large sample size of instances and tries to extract useful knowledge about the domain of parallel machine scheduling problem (PMSP) and solution space explored. The aim of this study is to provide statistical interpretations and classify the differences between the instances. The interrelationship between specified predetermined inputs and the output is examined through artificial neural networks (ANNs) along with regression analysis since they can easily explore which inputs are related to the output and develop regression model. The results of both analyses reveal significancies of the relationships and predicted importance of the predetermined inputs on the output. Furthermore, we examined the behaviours or patterns of the instances, after realizing the easiness and hardiness of the instances accentuating the differences. In order to link the predetermined inputs of instances with the performances of the set of tested methods, the differences between instances are evaluated in terms of variability. Then, we grouped instances into three clusters, specifying as exact, equal and difficult zones, for information retrieval about their complexities via hierarchical clustering method.