From Prediction to Action: Machine Learning and Optimisation for Better Decisions

Course overview

Data-driven decision-making relies on three pillars: descriptive analytics to understand what has happened, predictive analytics to anticipate what will happen, and prescriptive analytics to determine what should be done. While advances in machine learning have made descriptive and predictive tools widely accessible, many analysts still face the crucial question: “Now what?” How to turn predictions into optimal, real-world actions?

This hands-on course bridges this gap and introduces prescriptive analytics: the discipline of using data, models and explicit objectives to make optimal decisions. The two main tools of prescriptive analytics explored in this course are mathematical optimisation and machine learning. Participants will learn to frame a real problem in terms of decision variables, objectives, constraints and uncertain inputs, translate that structure into a mathematical model, implement it in Python, and turn the solution into a recommendation that can be evaluated and implemented.

This course is centred on practical modelling cases spanning business, finance, public policy and regulation, healthcare, energy, and logistics. The classical workflow in these cases comprises two steps: first, predict uncertain data; second, use the predictions to make better decisions. However, predictions contain errors, and not all errors have the same impact: overestimating no-shows on a half-empty flight has no consequences, but on a full flight it results in overbooking, upset customers, and compensation for damages. Therefore, predictions cannot be made in isolation, but must be integrated into the decision-making process. To this end, we develop an integrated approach in which machine learning models are embedded into mathematical optimisation models.

The course starts from the foundations of optimisation and builds towards integrated prediction-to-decision pipelines. The emphasis is on rigorous modelling, uncovering managerial insights and trade-offs, and practical use.