Multipopulation genetic programming for forecasting crop pests

This contribution attempts to study on forecasting crop pests with multipopulation genetic programming (MGP). In our previous work, standard genetic programming (SGP) evolves a single population, which often results in premature convergence. This paper concentrates on multipopulation evolution in order to maintain population diversity to avoid this. Comparison between single and multi population shows superiority of the latter. Study of migration interval and migration rate draws the conclusion that it is helpful to obtain optimal solutions that subpopulations keep communicating often and only a few of individuals migrate when communicating. All experiments are based on forecasting wheat stripe rust disease. MGP shows good prediction, which is hopeful to become an auxiliary method for forecasting crop pests.