Electronic Theses and Dissertation

Autism Spectrum Disorder (ASD) merupakan gangguan perkembangan saraf yang ditandai oleh
hambatan dalam interaksi sosial, komunikasi, serta pola perilaku yang terbatas dan berulang.
Variasi manifestasi klinis antar individu menyebabkan diagnosis ASD masih banyak bergantung
pada observasi subjektif. Penelitian ini bertujuan mengembangkan sistem klasifikasi
Electroencephalography (EEG) untuk membedakan subjek ASD dan normal menggunakan
Support Vector Machine dengan kernel Radial Basis Function (SVM-RBF) dan CatBoost. Data
EEG diperoleh dari 16 subjek yang terdiri atas 8 subjek ASD dan 8 subjek normal menggunakan
sistem EEG 16 kanal. Tahap prapemrosesan dilakukan menggunakan filter FIR berbasis jendela
Blackman–Harris pada rentang frekuensi 4–40 Hz. Sinyal kemudian disegmentasi menjadi
epoch berdurasi 2 detik pada frekuensi sampling 250 Hz, sehingga menghasilkan total 7.129
epoch yang terdiri atas 3.545 epoch dari kelompok ASD dan 3.584 epoch dari kelompok normal.
Selanjutnya dilakukan ekstraksi sebelas fitur temporal dan spektral sehingga diperoleh 176 fitur
pada setiap epoch. Seleksi fitur dilakukan menggunakan Mutual Information sebelum proses
klasifikasi. Evaluasi model menggunakan subject-wise stratified 8-fold cross-validation dengan
metrik accuracy, precision, recall, F1-score, dan specificity. Hasil penelitian menunjukkan
bahwa CatBoost memberikan performa terbaik dengan accuracy 99,56%, precision 99,72%,
recall 99,21%, F1-score 99,46%, dan specificity 99,72%, sedangkan SVM-RBF memperoleh
accuracy 91,35%, precision 88,15%, recall 95,35%, F1-score 91,60%, dan specificity 87,30%.
Uji paired t-test menunjukkan bahwa perbedaan performa kedua model signifikan secara
statistik (p < 0,001). Temuan ini menunjukkan bahwa kombinasi prapemrosesan Blackman
Harris FIR dan CatBoost merupakan pendekatan yang efektif dan andal untuk klasifikasi EEG
pada kasus ASD serta berpotensi mendukung pengembangan sistem bantu diagnosis yang lebih
objektif.
Kata kunci: Electroencephalography, Autism Spectrum Disorder, Blackman Harris, SVM RBF,
CatBoost

Autism Spectrum Disorder (ASD) is a neurodevelopmental disorder characterized by impairments in social interaction, communication, and restricted and repetitive patterns of behavior. The heterogeneity of clinical manifestations across individuals makes ASD diagnosis highly dependent on subjective observations. This study aimed to develop an Electroencephalography (EEG)-based classification system to distinguish individuals with ASD from normal subjects using Support Vector Machine with a Radial Basis Function kernel (SVM-RBF) and CatBoost classifiers. EEG data were acquired from 16 participants, comprising 8 individuals with ASD and 8 normal subjects, using a 16-channel EEG recording system. The preprocessing stage employed a Blackman–Harris finite impulse response (FIR) filter within the 4–40 Hz frequency range to suppress noise and unwanted frequency components. The filtered signals were segmented into 2-second epochs with 0% overlap, followed by the extraction of eleven temporal and spectral features, resulting in 176 features per epoch. Mutual Information-based feature selection was subsequently applied prior to classification. Model performance was evaluated using subject-wise stratified 8-fold cross validation based on accuracy, precision, recall, F1-score, and specificity. The experimental results demonstrated that CatBoost achieved the best performance, with an accuracy of 99.56%, precision of 99.72%, recall of 99.21%, F1-score of 99.46%, and specificity of 99.72%, whereas SVM-RBF achieved an accuracy of 91.35%, precision of 88.15%, recall of 95.35%, F1-score of 91.60%, and specificity of 87.30%. Furthermore, paired t-test analysis revealed that the performance differences between the two classifiers were statistically significant (p < 0.001). These findings indicate that the combination of Blackman–Harris FIR preprocessing and CatBoost classification provides an effective and reliable approach for EEG-based ASD classification and has the potential to support the development of more objective ASD diagnostic support systems. Keywords: Autism Spectrum Disorder, Electroencephalography, Blackman–Harris FIR, SVM RBF, CatBoost.