Comparison between fast AP-ANN and classical EP-ANN for lightning prediction / Azizi Ahmad Masduki

Lightning is an electrical discharge and produce the high energy which that brings at millions of volts and a few tens kilo ampere current. It is also produce the high temperature about thousand degrees Celsius within a few tens of milliseconds. Malaysia has high lightning occurrences because it is situated near the equator line which is characterized by high lightning activity. Over the years, various lightning prediction system have been developed and many technique have been presented to predict lightning. One of the methods for lightning prediction is by using an Artificial Neural Network (ANN) prediction system for lightning occurrences based on historical lightning and meteorological data from Malaysian environment. Using this method has a few problems about to finding suitable network architectures. This paper presented the improvement of method ANN with Evolutionary Programming (EP) as an optimization technique. This optimization technique will optimize to find ANN architectures systematically with less computation time. The mutations operators in EP discuss in this paper are Fast EP which apply Cauchy mutation and Classical EP which apply Gaussian mutation and the comparison for both of its. The best value sets of input data taken whether by using a Cauchy or Gaussian mutations and both operators will be compare to decide which the most suitable operators for lightning prediction is. As the result, the most suitable technique will create the best ANN architectures.