Demonstrating the Interply of Machine learning and optimization methods for operational planning Decision

Abstract In this paper, the interplaying of machine learning and optimization problems considered as indispensable problem-solving methods. While solving the real-world problem for distinctive aim but with the possibility of interaction, their integration is not as such a long history. Concerning this interplaying, literature summarized from the perspective of challenges in decision making and supply chain. Problem related to determining and forecasting both sales and demand in inventory decision problem based on classical methods is due to the underlying of relying on observational and censored data, which however is believed insufficient and mostly limited to upstream decision making. Inventory problem literature found encouraging in this regard especially related to incorporating side information for consistent decision-making. Paying attention to such side information at the qualitative scale, as the central scheme, and the classical inventory problem considered to predict then optimization demonstration of operational decision-making. With aim of bringing categorical scaled problems amenable to prediction, optimization and even to integration, encoding methods from machine learning and regressor compared at baseline. Consequently, information other than censored data, also become part of decision making process in this work to make decision consistent in one hand and ease at operational level (downstream) on the other hand. Moreover, the approach outperforms the classical state of an art method in terms of operating performance metrics and the cost found is stable.