Combinatorial optimisation problems arise in many areas of computer science and other disciplines, and typically involve finding groupings, orderings or assignments of discrete, finite sets of elements that satisfy certain conditions or constraints. Designing good optimisation algorithms for these problems is extremely difficult, and as such, automating the design and analysis of these algorithms is expected to have a very high impact across a range of applications. This work will explore if machine learning techniques can be leveraged to design good heuristics for graph-based optimisation problems. The learnt heuristics will be adapted for different input distributions.