Data Science meets Optimisation

There is an ever-growing interplay between data-science/machine-learning techniques and solving optimisation problems. 

Optimisation algorithms can benefit from data-science/machine learning to make better (internal) choices, and even to develop heuristics. This concept has been used in various areas of research, such as “hyper-heuristics”, “meta-optimisation”, “learning to search”, “algorithm selection”, “surrogate modelling”, etc.   

In a complementary fashion, machine learning is often in itself an optimisation problem, or requires optimisation, and so could potentially benefit from OR and AI optimisation techniques e.g., for “hyper-parameter tuning”, “algorithm selection”, “feature selection”, “data preprocessing”, “ensemble learning”, etc. 

This stream invites abstracts and presentations, either theory or practice-oriented exploiting the interaction between data science (machine learning, statistics, etc.) and optimisation that discuss any of (but not limited to) the issues of: 

  • Exact and inexact optimisation methods that are based on or improved/integrated with data-science and machine-learning techniques and/or allow a more general use of meta-heuristics and heuristics. For example, hyper-heuristics or adaptive search control systems. 
  • Machine learning methods that are based on, or improved/integrated with, exact and inexact optimisation techniques. 
  • Systems to build systems, particularly (meta/hyper-) heuristics or any other related (adaptive, self-tuning, reactive, self-improving, etc.) methods. 



Andrew Parkes, University of Nottingham


Ender Ozcan, University of Nottingham