Combining Artificial Bee Colony and Genetic Algorithms to Enhance the GPGPU- based ANN Classifier for Identifying Students with Learning Disabilities

摘要

Diagnosis of students with learning disabilities (LD) is a difficult procedure that requires extensive man power and takes a long time. Fortunately, through genetic-based (GA) parameters optimization, artificial neural network (ANN) classifier may be a good alternative to the above procedure. However, GA-based ANN model construction is computation-intensive and may take quite a while to process. In this study, we examine another optimization algorithm, the artificial bee colony (ABC) algorithm, which is based on the foraging behavior of honey bee swarm, to search for the appropriate parameters in constructing ANN-based LD classifier. We also integrate ABC algorithm with GA evolution strategy by first applying the former to derive a set of values of the ANN parameters and then use these values as the starting points for the latter GA evolution procedure. In addition, to speed-up the above process, a low-cost general purpose graphics processing unit (GPGPU), specifically, the nVidia graphics card, is adopted for the ANN model training and validation. The experimental results show that ABC can achieve better correct identification rate (CIR) than GA with less computation time. In addition, the strategy of using ABC as a pre-processing step for GA evolution has improved the correct identification rate by as much as 2.5% in two of our three data sets when compared to using GA alone.