Early Detection of Autism by Extracting Features: A Case Study in Bangladesh

Abstract

Autism Spectrum Disorder (ASD) is a neurobehavioral disorder that begins at childhood and exists this whole life. The objective of this work is that to explore significant features of normal and autism of divisional regions in Bangladesh. We collected individual samples of various children from their parents between 16 to 30 months of different residents using Autism Barta apps by web and fieldwork at Savar, Bangladesh. Then, we preprocessed our data and categorized frequent features based on their individual regions. Different tree based techniques such as J48, Logistic Model Tree, Random Forest, Reduced Error Pruned Tree, and Decision Stump were analyzed to find out the best classifier of them. From these classifiers, J48 showed the best outcomes than other classifiers. We extracted 9 rules and associated conditions from J48 decision tree and gathered frequent instances from our data for extracted rules. Finally, 8 within 23 features were required to classify normal and autism of individual regions in Bangladesh. Besides, 4 rules (10 Conditions) for normal and 5 (12 Conditions) rules for autism out of 9 (16 Conditions) rules were extracted from decision tree. This outcomes assist us to find out significant features of autism in Bangladesh. We expect that our work will be helpful things to improve their condition that leads them to a normal life.