Deteksi Gangguan Autis Pada Anak Menggunakan Metode Modified K-Nearst Neighbor

Abstract

ABSTRAK

Autism Spectrume Disorder merupakan ganggguan perkembanan dimana seorang anak memperlihatkan suatu perilaku menjauhkan diri dari lingkungan sosialnya, seakan – akan hidup di dunianya sendiri. Semakin dini mengetahui anak menderita gangguan autis sangatlahi penting, karenai semakini dinii autis diobati semakin besari peluang untuk sembuh. Dengani adanya perkembangani teknologii pendeteksiani gejalai autis dapati dilakukani lebih awal menggunakani sistem deteksi autis berbasis web. Metode Modified K-Nearest Neighbor merupakan perkembangan dari metode konvensional K-Nearest Neighbor. Deteksi autis pada anak menggunakan metode Modified K-Nearest Neighbor dengan data yang digunakan diambil dari website UCI Machine Learning Repository, dengan jumlah data 292 data dan 2 class output. Pengujian dilakukan dengan mencari kedekatan dari datai training dan data testing untuk menghitung Weight voting, setelah mendapatkan hasil weight voting akan dicari mayoritas datanya. Berdasarkani pengujiani yangi telah dilakukan didapatkan hasil akurasi tertinggi sebesar 96,67%, hasil presisi tertinggi sebesar 97,33% dan hasil recall tertinggi sebesar 100% pada K = 13. Untuk K optimal dari pengujian ini ditunjukkan pada K = 3.

Kata Kunci : Deteksi Autisme, Klasifikasi, Metode Modified K-Nearest Neighbor

ABSTRACT

Autism Spectrume Disorder is a development disturbance in which a child shows a behavior of distancing himself from his social environment, as though living in his own world. The earlier it is known that a child has an autistic disorder is very important, because the earlier autism is treated, the greater the chance of recovery. With the use of technological developments, autism symptom detection can be done earlier using a web-based autism detection system.Modified K-Nearest Neighbor method is a development of the conventional K-Nearest Neighbor method. Autism detection in children uses the Modified K-Nearest Neighbor method with the data used taken from the UCI Machine Learning Repository website, with a total of 292 data and 2 output classes. Testing is done by looking for the closeness of each training data to determine the validity value, after that look for the closeness of training data and testing data to calculate weight voting, after getting the results of weight voting the majority of the data will be searched. Based on the results of the test which have been done, it was found out that the highest accuracy results were 96.67%, highest precision results were 97,33% and highest recall results were 100% at K = 13. For the optimal k of this test was shown at K = 3.

Keywords: Autism Detection, Classification, Modified K-Nearest Neighbor Method.