<p>Discovering the hidden relationships between the electricity market dynamics and the power system security represents a strategic leverage for enabling power system planning and operation tools to assess the dynamic grid impacts of new market schemes, different generation mixes, and available portfolio technologies. Moreover, it allows inferring the more relevant variables and the corresponding maximum/minimum thresholds that could threaten the dynamic power system security, hence enhancing the situational awareness of the transmission system operator with prompt, reliable and actionable intelligence. To address this challenging issue, this paper advocates the role of probability-based feature selection and machine learning techniques for processing the energy market outcomes and directly classifying the corresponding grid security states. The main idea is to identify the most promising techniques that enable learning, from a set of historical observations, the unknown mapping between the market outcomes and the corresponding dynamic security state computed by a real Dynamic Security Assessment tool. Detailed experimental results obtained on a synthetic one-year operation scenario defined by the Italian transmission system operator are presented and discussed to outline the most promising enabling methodologies.</p>