Multi-objective evolutionary beer optimisation

Abstract

Food production is a complex process which can benefit from many optimisation\napproaches. However, there is growing interest in methods that support\ncustomisation of food properties to satisfy individual consumer preferences.\nThis paper addresses the personalisation of beer properties. Having identified\ncomponents of the production process for craft beers whose production tends to\nbe less standardised, we introduce a system which enables brewers to map the\ndesired beer properties into ingredients dosage and combination. Previously\nexplored approaches include direct use of structural equations as well as\nglobal machine learning methods. We introduce a framework which uses an\nevolutionary method supporting multi-objective optimisation. This work\nidentifies problem-dependent objectives, their associations, and proposes a\nworkflow to automate the discovery of multiple novel recipes based on\nuser-defined criteria. The quality of the solutions generated by the\nmulti-objective optimiser is compared against solutions from multiple runs of\nthe method, and those of a single objective evolutionary technique. This\ncomparison provides a road-map allowing the users to choose among more varied\noptions or to fine-tune one of the favourite identified solution. The\nexperiments presented here demonstrate the usability of the framework as well\nas the transparency of its criteria.\n