Application of ANFIS-based Non-Linear Regression Modelling to Predict Concentration Level in Concentration Grid Test as Early Detection of ADHD in Children

Abstract

Concentration is the main asset for students and serves as an indicator of successful learning implementation. One of the abnormal disturbances that can occur in a child's concentration development is attention deficit hyperactivity disorder (ADHD). The prevalence of ADHD in Indonesia in 2014 reached 12.81 million people due to delayed management in addressing ADHD. Therefore, early detection of ADHD is necessary for prevention. ADHD detection can be done by testing the level of concentration using a concentration grid. However, a method is needed that can be applied to uncooperative young children who are not familiar with numbers. Therefore, research was conducted with an innovative approach using a combination of EEG-ECG to classify concentration levels. The data used in this study were primary data from 4 participants with 5 repetitions. The data were processed in the preprocessing stage, which involved noise filtering and Butterworth filtering. The features used in this study were BPM (beats per minute), alpha, theta, and beta EEG signals, which would later become inputs for the Adaptive Neuro-Fuzzy Inference System (ANFIS). The output shows that the combination of EEG-ECG has the potential to predict concentration test results. Using BPM, alpha, theta, and beta signals can serve as parameters for predicting the concentration grid test values using ANFIS effectively. In the ANFIS model with 4 features, an accuracy of 99.997% was obtained for the training data and 80.2142% for the testing data. This result could be developed for early detection of ADHD based on concentration levels so the learning implementation could be more effective.