Employing Machine Learning-Based Predictive Analytical Approaches to Classify Autism Spectrum Disorder Types

Table 1

Summary of related work.

Sr. No.	Authors	Objectives	Method	Dataset	Accuracy
1	Dvornek et al [22]	Used the long short-term memory network to identify the autism from resting-state fMRI.	Recurrent neural networks (long short-term memory)	ABIDE	68.5%
2	Van den Bekerom [23]	To use the machine learning to predict the ASD.	Machine learning algorithms (SVM, RF, and NB) and 1-way method	National Survey of Children’s Health (NSCH) data	0.49% to 0.54% in 4 classes using machine learning and 54.1% to 90.2% using the 1-way method
3	Heinsfeld et al [24]	To apply the deep learning algorithm to identify the ASD.	SVM, RF, DNN	ABIDE	65%, 63%, 70%
4	Bi et al [25]	To apply the random SVM to classify the ASD.	Random support vector cluster	ABIDE	96.15%
5	Altay and Ulas (2018)	To use the classification method to identify the children who have ASD or not.	Discriminant analysis, KNN	292 samples	91%, 89%
6	Kong et al [28]	To identify the ASD using DNN.	DNN	ABIDE	90.39%
7	Eslami and saeed [29]	To diagnose the ASD using the Auto-ASD-Network based on the deep learning and SVM.	DNN, SVM, state-of-the-art classifier	4 fMRI datasets	80%
8	Wilson and rajan [30]	To use the brain imaging dataset to identify the ASD.	DL algorithms	ABIDE	70%
9	Sudha and Vijaya [31]	To identify the ASD by supervised algorithms.	Decision tree, SVM, MLP		98%, 96%, 95%
10	Nasser et al [32]	To identify the ASD by artificial neural network.	ANN	Dataset collected from the ASD screening app	100%