Predicting Geomagnetic Activity Index by Brain Emotional Learning ALI GHOLIPOUR † , CARO LUCAS † , DANIAL SHAHMIRZADI ‡ † Control and Intelligent Processing Center of Excellence, Electrical and Computer Eng. Department, University of Tehran, Tehran, Iran And School of Intelligent Systems, Institute for studies in theoretical Physics and Mathematics, Tehran, Iran ‡ Mechanical Engineering Department, Texas A&M University, College Station, Texas Abstract - The Emotional Learning Algorithm has been introduced to show the effect of emotions as well known stimuli in the quick and almost satisficing decision making in human. The remarkable properties of emotional learning, low computational complexity and fast training, and its simplicity in multi objective problems has made it a powerful methodology in real time control and decision systems, where the gradient based methods and evolutionary algorithms are hard to be used due to their high computational complexity. Recently the emotional approach has been successfully used to obtain multiple objectives in prediction problems of real world phenomena, more specifically space weather forecasting. A newer and more realistic model of emotional learning in human brain is used in this study to predict the most popular index of geomagnetic activity: Kp. The interplanetary Kp index is mainly used in warning and alert systems for satellites. The proposed model with brain emotional learning algorithm is introduced to make purposeful prediction of Kp index. Both the prediction accuracy during geomagnetic disturbances and sub- storms, and the rate of associated correct warning messages show the efficiency of this algorithm. Key Words: - Prediction, Emotional Learning, BEL, Multi Objective Decision Making, Space Weather, Geomagnetic Activity, Warning System. 1. Introduction The solar wind driven magnetosphere is a complex dynamical system with highly nonlinear and chaotic behavior [1]. A large number of studies have been carried out to provide appropriate dynamical models of magnetosphere, and to predict various geomagnetic indices, e.g. Dst storm time index and AE auroral electrojet index [2,3,4,5]. But the most popular indicator of geomagnetic disturbances, the Kp index, which is used mainly in warning and alert systems for satellites, has not been considered as much. The planetary geomagnetic index, Kp, expresses the average amplitude of the horizontal components obtained from 13 selected observatories situated at mid latitudes between 45 and 60 degrees [6]. It is a quasi logarithmic scale in 3 hour intervals from 0 to 9 in thirds of a unit (totally 27 discrete values): 0 0 0 0 9 , 9 , 8 , 8 , 8 , , 1 , 1 , 1 , 0 , 0 − + − + − + K where 0 means field amplitude less than 5 nT and 9 shows a geomagnetic activity more than 500 nT. Kp index is a good monitor for the warning and alert systems for satellites. It is used as a surrogate for short term atmosphere heating caused by geomagnetic disturbances. The spacecraft drag, latitude losing in low earth orbit satellites, is an important problem as a result of atmospheric heating, which occurs normally at Kp values more than 6. Some spacecrafts use the earth’s magnetic field as an aid in orientation, or as a force to dump momentum. The geomagnetic disturbances with Kp values exceeding 5 normally cause miss- orientations. Finally, surface charging effects at Kp values between 4 and 5 have long term effects on satellites. Considering all of these, predicting the Kp index is very important for geomagnetic K-index warning and alert systems. There are four warning messages for Kp which are issued or extended for any period with expected values of Kp = 4, 5, 6, or >= 7. Higher index warnings supersede lower ones. In this study, a recently developed model of emotional learning in human brain [7,8] is considered to be used in purposeful prediction of Kp index and to provide a more reliable K-warning system. The remarkable properties of emotional learning, low computational complexity and fast training, and its simplicity in multi objective problems has made it a powerful methodology in real time control and decision systems, where the gradient based methods and evolutionary algorithms are hard to be used due to their high computational complexity [9,10,11,12,13,14]. While most of previous studies satisfy the objectives via fuzzy implementations of emotional critics [15,16,17], the simulated emotional learning procedure [7,8] inherently emphasizes to learn the features related to high values of inputs considered as high level stimuli to the brain