Enhancing Special Needs Identification for Children: A Comparative Study on Classification Methods Using ID3 Algorithm and Alternative Approaches

Abstract

This research aims to compare the performance of classification methods in identifying special needs in children. The dataset used consists of identifications of various types of special needs, such as ADHD, autism, mild cerebral palsy, mild intellectual disability, moderate intellectual disability, and hearing impairment. The methods compared include ID3 (previous study), Naive Bayes, Random Forest, k-NN, and Gradient Boosting. The comparison results show that ID3 achieves an accuracy rate of 91.81%. The new alternative methods show better performance, with Naive Bayes achieving an accuracy of 95.28%, Random Forest 95.14%, k-NN 95.28%, and Gradient Boosting 83.47%. Although Random Forest does not outperform Naive Bayes and k-NN, it has the advantage of forming decision trees that align with symptom attributes and predict disability labels. However, in the implementation of the Gradient Boosting algorithm, there is a low model probability, especially in identifying ADHD. The conclusion of this research provides insights for researchers in selecting appropriate classification methods for identifying special needs in children, considering accuracy, efficiency, and handling imbalanced data.