Autism spectrum disorder (ASD) is growing faster than ever before. Autism detection is costly and time intensive with screening procedures. Autism can be detected at an early stage by the development of artificial intelligence and machine learning (ML). While a number of experiments using many approaches were conducted, these studies provided no conclusion as to the prediction of autism characteristics in various age groups. This chapter is therefore intended to suggest an accurate MLASD predictive model based on the ML methodology to prevent ASD for people of all ages. It is a method for prediction. This survey was conducted to develop and assess ASD prediction in an artificial neural network (ANN). AQ-10 data collection was used to test the proposed pattern. The findings of the evaluation reveal that the proposed prediction model has improved results in terms of consistency, specificity, sensitivity, and dataset accuracy.
TopIntroduction
Diagnoses of ASD based on behavioural characteristics (Omar et al., 2019), although in many but most cases, the exact cause of ASD is not understood (Cruz & Wishart, 2006). Reliable determination of ASD is a big problem for ASD monitoring schemes and broad studies of ASD generally.
Although stringent ASD diagnostic instruments do exist, clinicians use a range of daily practical resources and approaches (Khan et al., 2017). It is also impossible to identify ASD using “gold-standard” clinic-based practises for large-scale or population-based trials. Many trials in epidemiology are based somewhat, and often only, on current “administrative” ASD grading designations: ICD 9, billing codes, special education or the autism related disorder insurance eligibility (e.g., Medicaid) (Wall et al., 2012; Bone et al., 2016). There is a substantial inconsistency in the use of these classifications in the United States. These programmes are not universal in determining all persons that follow the population ASD requirements and their main goals are ensuring adequate provision of care to people rather than classifying disabilities (Allison et al., 2012; Thabtah, 2017).
The data used to test atypical variances in children with and without ASD in face expressions (Hauck and Kliewer, 2017). Using knowledge theory, mathematical analysis and time series simulation, six facial responses were securitised. Researchers have recently embraced data analytics as a diagnostic method for ASD in refinement and implementation. The possibility of grouping in the different domains has been revealed by being an effective computer tool (Bekerom, 2017). Therefore, some literature has used machine learning techniques in order to distinguish persons with and without ASD that are discriminating against neurons (Wall, 2012), behaviour indicators (Heinsfeld et al., 2018) and (Saini et al., 2021) sentiments respectively. The effect of the eyes on children with high- or low-risk children at high-ASD levels. Computing approaches were applied, such as 0.61 accurate discrimination analysis for the discriminating functions, 0.64 precision support for vector machine, and a 0.56 accuracy linear discrimination analysis. The Autism-diagnostic interview and the social responsiveness level were revised, SMV machine learning categories were extended to a leading sample of people with or without ASD, and positive findings were obtained (Bone et al., 2015).
(Kosmicki et al., 2015) used artificial neural networks to create a predictive model based on a data collection of 22 possible risk elements for pregnancy in the development of autism (Omar et al., 2019). The predictive accuracy of the artificial neural network was 83 percent. Their research promoted the use of ANNs as an outstanding diagnosis monitoring method for ASD, and welcomed them. The highest level of accuracy achieved to date is 95.07 percent for SVM (Reyana et al, 2020).
Autism diagnosis takes a great deal of time and expenses. Earlier diagnosis of autism behaviours, such as verbal, nonverbal speech, perseverance, physical skills, repeated behaviour, sensory treatment and social awareness etc. may be helpful when administering early medicine to patients, as seen in Figure 1. It may help prevent further deterioration of the health of the patient, and reduce the long-term costs of late diagnosis. Autism characteristics in a person and whether or not an extensive autism evaluation is needed.
This research aims to formulate a model of autism prediction using ML techniques and create a smartphone app that can accurately predict the autism traits of a person of any generation. This thesis focuses on the development of an autism screening application to predict ASD characteristics of individuals between the ages of 4-11 years, 12-17 years and 18 years and older.
The key problem with a perfect SMM recognition is the identification of the most significant characteristics which carry stereotypical behaviour by means of an automated feature extraction technique from accelerometers. Another problem is personalization caused by variability within and between subjects (Liaw and Wiener, 2002). Intrasubject variances can in fact be explained by variations in severity, length and frequency of SMM in any atypical subject (subject of the automotive spectrums) whereas variances between SMMs among different atypical subjects can be explained by inter-subject variances (Bone, 2015).
An adaptive method (Sampathkumar et al, 2020) is thus essential in order to generalise all SMMs through individuals and to respond to new SMM behaviour. These problems can be solved with profound SMM identification learning strategies. We train two CNN models inside the subjects in times and frequency areas whose parameters are selected on the basis of the properties of the SMM input.