The process of steel casting involves several energy-intensive tasks such as heat transfer, solidification process, etc. Though the evolution of continuous casting of steel from the conventional ingot casting enabled a high amount of energy savings, operational parameter optimization considering various avenues of uncertainty is the key for next level of improvement in energy efficiency and process sustainability. To achieve this, a multi-objective optimization formulation under uncertainty has been proposed that can lead to simultaneous maximization of productivity and minimization of energy consumption. Among various uncertainty handling techniques, Chance Constrained Programming (CCP) is considered as an efficient approach. However, the requirement of uncertain parameters to follow some well-behaved probability distribution for having a closed form analytical solution in CCP is a bottleneck for most of the practical situations due to the unknown nature of uncertain data. This paper proposes a novel methodology called DDCCP (Data-Driven CCP), to amalgamate machine learning algorithms with CCP, thereby making the approach data-driven. A novel fuzzy clustering mechanism is implemented to transcript uncertain space such that the specific regions of uncertainty are identified accurately based on given uncertain data for more realistic sampling and thereby impacting the optimal solution accuracy. Implementing DDCCP on the casting model, ∼20–70% improvement in the objectives of energy calculations and 50–100% improvement in the metrics of Pareto optimal solutions are observed as compared to the existing box sampling approach showing efficacy of the proposed methodology. © 2019 Elsevier Ltd