Electronic Theses and Dissertation

Autism Spectrum Disorder (ASD) merupakan gangguan perkembangan saraf yang memengaruhi kemampuan komunikasi, interaksi sosial, dan perilaku individu. Deteksi dini ASD penting untuk mendukung intervensi yang lebih efektif, namun diagnosis konvensional masih bersifat subjektif dan memakan waktu. Penelitian ini bertujuan menganalisis performa klasifikasi ASD berbasis citra wajah melalui ekstraksi fitur jarak antar titik facial landmark menggunakan algoritma machine learning. Dataset terdiri dari 2.032 citra wajah anak (ASD dan Non-ASD) yang diproses melalui deteksi wajah, penandaan 68 titik landmark, serta perhitungan jarak Euclidean sebagai fitur geometris. Dua algoritma klasifikasi, Logistic Regression dan Extra Trees Classifier, diterapkan untuk membedakan karakteristik wajah kedua kelompok. Evaluasi dilakukan menggunakan akurasi, presisi, recall, dan F1-score. Hasil penelitian diharapkan menghasilkan model klasifikasi dengan akurasi yang lebih tinggi serta mendukung pengembangan sistem deteksi dini ASD berbasis web yang objektif dan efisien.

Autism Spectrum Disorder (ASD) is a neurodevelopmental disorder that affects communication skills, social interactions, and individual behavior. Early detection of ASD is crucial for more effective interventions, but conventional diagnosis remains subjective and time-consuming. This study aims to analyze the performance of facial image-based ASD classification through feature extraction of distances between facial landmarks using a machine learning algorithm. The dataset consists of 2,032 facial images of children (ASD and non-ASD) processed through face detection, marking of 68 landmark points, and calculating Euclidean distance as a geometric feature. Two classification algorithms, Logistic Regression and Extra Trees Classifier, were applied to differentiate the facial characteristics of the two groups. Evaluation was carried out using accuracy, precision, recall, and F1-score. The results of this study are expected to produce a classification model with higher accuracy and support the development of an objective and efficient web-based ASD early detection system.