Price optimisation of perishable goods using a genetic algorithm

Multi-product profit optimisation problems have been studied under nested logit models of consumer behaviour. Although attractive through to the relaxation of strong assumptions of multinomial logit models, nested logit models as well as multinomial logit models require costly discrete choice experiments in order to collect data for estimating model parameters. We propose a novel formulation of multi-product profit optimisation that is especially useful for perishable goods that are of the same type and different only in their quality level. Our model relies on willingness to pay data that can be elicited directly, derived from market data or measured indirectly in auctions or through transactions. We furthermore present a genetic algorithm for solving the formulated multi-product profit optimisation and show that our proposed genetic algorithm finds nearby optimal solutions within a very short time span.