Uncertainty in process conditions makes the optimal solution from the deterministic optimization case infeasible or suboptimal. This mandates a special treatment for process uncertainty. The advent of unsupervised machine learning technique like clustering can be used to address this problem. Data arising from process conditions is meager and, to perform robust optimization, it is necessary to have enough representation of the uncertain region. For this, the collected data from process conditions is clustered using Support Vector Clustering (SVC). The hyper-parameters of SVC play a crucial role in clustering exercise. To promote this, a novel algorithm for determining optimal SVC models, is proposed. The proposed algorithm is used to cluster the uncertain process data and once clustered, can be used to sample additional data points inside each cluster using the convex hull. The additionally sampled data points in each cluster increases the cluster representativeness. The proposed methodology is implemented on industrial grinding circuits for performing Optimization Under Uncertainty (OUU) for the worst-case scenario and tested against the budget-uncertainty set based OUU. © 2022 IEEE.