Hybrid evolutionary programming-artificial neural network-based lightning prediction system / Dalina Johari

This project presents the development of hybrid evolutionary programming (EP) and artificial neural network (ANN) prediction system for lightning occurrence based on historical lightning and meteorological data from a Malaysian environment. It involved the development of ANN design and embedding EP optimization technique for optimizing selected ANN parameters in order to improve the system's generalization capability.ANN, an intelligent machine learning technique, is inspired by the way our biological nervous systems process information. With the ability to learn by example and do tasks based on training experience, it is profoundly suitable for pattern recognition and forecasting tasks. Due to ANN heuristics nature, though, the process of finding suitable network architectures could be arduous and time-consuming. It highly depends on expert experience and is a tedious trial and error process. There is also no systematic way to design the optimal architecture for a given task automatically. For that reason, an efficient optimization technique such as EP was employed in the study to find the best ANN architectures systematically and perform lightning prediction accurately with less computation time. Comparative study conducted between determining ANN parameters heuristically and by using EP revealed that hybrid EP optimization technique with the ANN produced better results for the ANN in terms of its R-value and computational time. The most significant advantage of using EP optimization technique is that it provides a structured and automatic way for obtaining optimal values of the ANN parameters, while using heuristic technique, each possible value had to be tried one by one manually before the optimal values could be found. As a result, the developed lightning prediction system is able to generalize well when presented with new sets of input data. Consequently, prediction of the lightning occurrence can be successfully done.