Classification of Autism Spectrum Disorder (ASD) Using Multi-Layer Perceptron (MLP) Neural Network

Abstract

Autism Spectrum Disorder (ASD) is a neuro-developmental condition characterized by developmental disability, speech and language delays, abnormal social interactions, behaviour excesses and repetitive and stereotyped behavior. This research is aimed at improving the early diagnosis of ASD using deep learning technique. Raw datasets related to both Autism and Non-Autism cases were collected from the University of Califona Irvine (UCI ) repository. Multi-layer perceptron (MLP) neural network was used for the classification. Keras was used to build the MLP neural network model with the aid of Jupyter Notebook. Some of the metrics used to evaluate the model are performance accuracy, probability, roc-auc curve and confusion matrix which shows the precision, recall and F1 score. After running 7 epochs, 98.3% performance accuracy was achieved for the training data while 98.1% performance accuracy was achieved for the test data. With higher epochs, better results can still be achieved. This shows that MLP performs very well in classifying ASD dataset, making it easy for early diagnosis of the disease. Keywords: Autism spectrum disorder, Classification, Machine learning, MLP. Ayorinde I. T. & Bankole O. A. (2023): Classification of Autism Spectrum Disorder Using Multi-Layer Perceptron Neural Network. Journal of Advances in Mathematical & Computational Science. Vol. 11, No. 3. Pp 77-88. dx.doi.org/10.22624/AIMS/MATHS/V11N3P5. Available online at www.isteams.net/mathematics-computationaljournal.