A machine learning-based framework for data mining and optimization of a production system

In the present paper, we performed several decision tree algorithms to classify instances and represent the most efficient policies depicted by a hybrid reinforcement learning algorithm and treat a complex production, maintenance and quality control optimization problem within a degrading manufacturing and remanufacturing system. The constructed decision trees contained of nodes, which represent its independent variables, and leaves that stand for the set of function values. All optimization parameters and optimal policies found by the hybrid reinforcement learning algorithm, used as the training set for the trees algorithms. After the construction of each tree, the resulting rule used to treat the optimization problem and the performance of each rule compared. In addition, for the best performing trees algorithms, further investigation performed for the impact of their parameters to its rule effectivity.