Genetic Programming VS Statistical Modelling for ecological data prediction

West Nile Virus (WNV) is an emerging global health threat. Transmission risk is strongly related to the abundance of mosquito vectors, typically Culex pipiens in Europe. Data on Culex pipiens mosquito captures from the Piedmont region were analysed to determine the principal drivers of mosquito population dynamics. The main goal was to build predictive models to help in controlling the abundance of mosquitoes. Two different approaches were implemented and compared: bayesian Generalized Linear Mixed Model (GLMM), i.e. a statistical model, and genetic programming. Genetic programming is an evolutionary computational technique that automatically solves problems without requiring a-priori specification of the structure of the solution. It has been successfully used in various applications such as data modelling, design creation, signal and image processing, industrial process control. The comparison between the two methods was performed using the Root Mean Square Error (RMSE). Genetic programming achieved overall better performance than the statistical model and it was able to automatically capture interactions between features. Furthermore, solutions returned by genetic programming are explicit and use a rather limited subset of the available variables.