Autistic Spectrum Disorder (ASD) is one of the neurodevelopment conditions associated with significant healthcare costs. Examples of ASD traits include problems with social interaction with others, unusual interest in objects, need for sameness, under- or over-reaction to one or more of the five senses (sight, touch, taste, smell, or hearing) and unusual emotional reactions and expressions. Although there is no cure for ASD, early diagnosis combined with appropriate treatments such as therapies can reduce the symptoms and improve learning and development. However, the waiting time for an ASD diagnosis is usually lengthy, making a lot of ASD patients fail to receive appropriate early treatments. As a result, they might have already developed irreversible developmental problems where more expensive and intensive treatments are required. The economic impact of autism and the increase in the number of ASD cases across the world reveals an urgent need for the development of easily implemented and effective screening methods. Dr Thabtah from the Manukau Institute of Technology developed a mobile app called ASDTests that implements ASD screening methods. In the app, different questionnaires are designed for different user groups, including toddlers (less than or equal to 36 months), children (4 11 years), adolescents (12 16 years), and adults (17 years and above). The dataset (ASD_toddler.csv) contains the screening records belonging to toddlers. The descriptions of the fields in the dataset are listed in Table 1. The questionnaire designed for toddlers consists of 10 questions (A1-A10) focusing on behaviours that reflect autistic traits in very early childhood. Each question has five response options based on the frequency to which the toddler exhibits the behaviour. The response to each question is converted into a binary value. If a particular autistic behaviour is not frequently observed, then the value of 0 will be assigned. How the responses to A1-A10 are converted into binary values can be referred to the third column of Table 1. Note that A10 is reverse-scored.

In this question, you are going to import the dataset to the IBM SPSS Modeler and perform all data mining tasks (including data preparation, if any) using the IBM SPSS Modeler. (a) Based on the responses to the questionnaire, a screening score can be calculated by summing the values of all the questions from A1 to A10. A high screening score suggested that a toddler might have ASD traits and should pursue a formal clinical diagnosis. Derive a new field Screening_Score using IBM SPSS Modeler accordingly. Then, report the mode of this new field. Provide the necessary screenshot(s) to support your answer

(b) According to domain experts, a screening score greater than 3 denotes a clinical assessment is necessary; otherwise, no autistic traits are found. Use the dataset prepared from part (a) to perform data visualisation in order to answer the following enquiries:

Enquiry 1: How many toddlers are suggested to pursue clinical assessment of ASD?

Enquiry 2: What are the top three autistic behaviours (i.e., A1, A2, , A10) observed among toddlers who need clinical assessment of ASD? Provide one (1) graphic display for each enquiry to support your answers. If deemed suitable, you could further prepare the dataset using the IBM SPSS Modeler before generating the graphs. If you do this, provide the necessary screenshot(s) and explain.

(c) Given that the data mining objective is to identify the autistic behaviours (i.e., A1, A2, , A10) frequently co-exhibited by toddlers who need clinical assessment of ASD, construct an association rule mining model using the Apriori algorithm. Discuss how the association rules obtained can help health professionals in the context of ASD. (Note: Pick at least two rules that you find interesting to discuss.) In your answer, provide screenshots that show clearly

(i) which fields are selected for the analysis and their roles,

(ii) the parameters used in the algorithm, and

(iii) the association rules that you used for discussion. (d) Both web graphs and association rules can help identify the co-occurrences between categorical fields. Differentiate web graphs and association rules by suggesting two (2) differences between them.

Table 1. Description of the dataset ("ASD toddler.csv")