When mushrooms are produced in commercial quantities, the quality and yield of the mushroom crop can be seriously damaged
through infestation by sciarid flies. An important weapon in combatting sciarid fly is the use of nematode worms, Steinernema
feltiae, which are sold commercially for bio-control on mushroom farms.
In this work, we develop optimised control strategies, using a Markov Network EDA to decide whether a fixed dose of nematodes
is either applied or not applied at each of a series of potential intervention points during the period of mushroom cultivation.
There is a wider motivation here which is to explore the relationship between optimisation and the probabilistic model built
by an EDA. We have shown that coefficients of the probabilistic model generated by the EDA can be understood in terms of the
real-life problem, in particular relating to the larval phase of the sciarid fly life cycle, when most damage s done to the crop.
Our ongoing interest in this area is to use probabilistic models to detect points of sensitivity to intervention in
complex non-linear control systems. We also intend to develop evolutionary optimisation approaches to exploit this information
to derive low-cost, highly effective intervention strategies.