2.1 Identify the issues that you will investigate

[sapphirachan]

Q-Chat-10 is a screening method that consist of only 10 questions, with an equal weightage on the on the answers for all questions, and with a low threshold of testing as positive for ASD with just 3 out of any 10 questions answered. We want to investigate if there should be different weightage to be assigned to questions to which are linked to behavioural traits that are more prominent based on our analysis, so as to improve the accuracy of the screening.

We aim to investigate the following questions:

• Identify the most influential and significant questions in Q-CHAT-10 for ASD screening
• Identify the most suitable machine learning methods to detect ASD in toddlers
• Determine if current testing methods of Q-Chat can be improved

[flfguerrero]

ASD screening in toddlers is currently a user operated task performed by caretakers (parent, guardians, teachers etc.) given a set of observable behavioural characteristics (communication/interactions) display themselves which raises concern. These screening methods, required manual input from users which can be inefficient (slow) and inaccurate (human error), but also fails to take into account the impactful and significant questions that can imply ASD.

In order to address these issues and provide proper ASD screening in toddlers. We propose to utilise data mining techniques and machine learning. This, however, is a largely unexplored research in a field of expertise that requires extensive scientific and medical knowledge.