Auto-bias selection for developing learning-based scheduling systems

Inductive learning techniques have shown encouraging achievements in reducing the effort for knowledge acquisition in the development of knowledge-based scheduling systems. However, there is little research on selecting the proper learning biases to facilitate the development of knowledge bases (KBs), so that the prediction accuracy of the resulting KBs is enhanced. In this study, we propose an auto-bias selection procedure that is capable of determining good learning biases, such as a proper feature set (system attributes) and a suitable learning algorithm. Through the case study conducted, the proposed procedure shows its capability to select proper learning biases and the results of experiments demonstrate that the scheduling KBs, developed by the learning biases selected, have a superior ability to produce the correct scheduling strategy over the scheduling KBs developed without exploring the proper learning biases.