Identifying Optimal Feature Set for Improved Autism Classification Using Machine Learning Techniques
C Karpagam, C. Deepa · International Journal of Engineering Trends and Technology · 2024
Administering standard medical prognosis tools for autism disorder is a time-consuming process. Furthermore, only a trained and experienced professional can supervise the assessment. The attempts that failed to evaluate ASD (Autism Spectrum Disorder) at the right time lead to critical medical care costs and a high impact on an individual’s performance in regular activities. More flexible and evident accessible methods would assist parents and caretakers in mitigating the hurdles faced during conventional clinical diagnosis. The previous work represents the exploratory data analysis made on the autism dataset. Here, an expansion to build a model by combining Machine Learning classification algorithms on selected feature sets for improved accuracy is administered. The autism dataset used in the experiment is collected from a public repository that includes 1054 instances. RFE (Recursive Feature Elimination) and the Boruta method are preferred to determine the relevant feature set with the highest rank. A significant improvement in the accuracy of results is noted when a Random Forest (RF) is ensembled with a Support Vector Machine (SVM) with 98.97% accuracy in the toddler dataset. The resultant model maximizes accuracy and minimizes the efforts taken by practitioners with the extensive diagnosis process.