Differential evolution tuned support vector machine for autistic spectrum disorder diagnosis

Abstract

Autistic Spectrum Disorder (ASD) is a brain developmental disorder which weakens the ability to communicate and interact with others. A child with autism spectrum disorder may have different, repetitive patterns of behaviour, interests or activities, including some specific signs. To diagnose the behaviour of ASD and identify the level of disease on the human is still a challenging task for the doctors. Only by the trained and experienced physician can identify the ASD immediately. The data set for autism problem consist of number of causes and the results based on the symptoms for ASD. So, Data mining algorithm is in need to organize and pattern the ASD details. The machine algorithms are available to classify the data in data mining works. In this proposed work, a machine learning algorithm called Support Vector Machine is used to classify the ASD children accurately. SVM is one of the classification algorithms which finding the hyper plane that maximizes the margin between the two classes. Though SVM give better identification of disease, some children have their unique nature which hides their problem of ASD easily. So, to diagnose the problem accurately, the user defined SVM parameters are tuned by optimization algorithm called Differential Evolutionary Algorithm. DE is an optimization algorithm used to find the optimal solution of SVM parameters. Further, to improve the performance of the proposed method, the dimension reduction technique is followed to reduce the SVM and ANN network dimension. The Sequential Feature Selection (SFS) method is applied in this paper, which select the most influenced variables for the output. The reduced network is further classified by ANN and SVM model. The Data set for the ANN and SVM network has been taken from the real records of the multi-specialty hospitals. The SVM and DE optimized SVM results are compared with another classification model called Artificial Neural Networks. The test results show the betterment of DE optimized SVM which give the classification of ASD child very accurately compare with ANN and DE optimized ANN.