Jurnal Ilmiah Kursor
Vol 6 No 4 (2012)

SOLAR STORM TYPE CLASSIFICATION USING PROBABILISTIC NEURAL NETWORK COMPARED WITH THE SELF-ORGANIZING MAP

Gregorius S. Budhi (Information Engineering Department, Petra Christian University)
Rudy Adipranata (Information Engineering Department, Petra Christian University)
Bambang Setiahadi (Aerospace Observer Station (SPD) LAPAN)
Adrian H. N (Information Engineering Department, Petra Christian University)



Article Info

Publish Date
23 Jul 2012

Abstract

Salah satu tugas LAPAN adalah melakukan pengamatan dan peramalan gangguan badai matahari. Gangguan ini dapat mempengaruhi medan elektromagnet bumi sehingga mengacaukan peralatan elektronik dan navigasi yang ada di bumi. Hal ini akan dapat membahayakan kehidupan manusia bila tidak diantisipasi dengan benar. LAPAN menginginkan adanya aplikasi komputer yang secara otomatis dapat mengklasifikasi tipe badai matahari, yang menjadi bagian dari sistem early warning yang akan dibuat. Oleh sebab itu kami dari Universitas Kristen Petra program studi Teknik Informatika dan Lembaga Penerbangan dan Antariksa Nasional melakukan penelitian bersama tentang klasifikasi badai matahari. Klasifikasi dilakukan terhadap citra digital badai matahari / kelompok bintik matahari ini berbasis pada sistem klasifikasi ”Modified - Zurich Sunspot Classification” yang banyak digunakan. Metode klasifikasi yang kami gunakan disini adalah Jaringan Saraf Tiruan Probabilistik. Hasil dari pengujian cukup menjanjikan karena memiliki akurasi sebesar 94% untuk data testing. Akurasi ini lebih baik dari akurasi aplikasi serupa yang dibangun dengan kombinasi metode Self-Organizing Map dan K-Nearest Neighbor. Kata kunci: Klasifikasi Tipe Badai Matahari, Sistem Klasifikasi Modified - Zurich Sunspot Classification, Jaringan Syaraf Tiruan Probabilistik. Abstract One of the task of the LAPAN is making obsevation and forecasting of solar storms disturbance. This disturbances can affect the earth's electromagnetic field that disrupt the electronic and navigational equipment on earth. It would be dangerous to human life if not properly anticipated. LAPAN wanted a computer application that can automatically classify the type of solar storms, which became part of early warning systems to be created. Therefore we from Petra Christian University Informatics Engineering Department and the Indonesian National Aeronautics and Space Agency conduct joint research on the classification of solar storms. The classification of the digital images of solar storm / sunspot groups is based on “Modified - Zurich Sunspot Classification System” which is widely used. Classification method that we use here is the Probabilistic Neural Networks. The result of testing is promising because it has an accuracy of 94% for testing data. The accuracy is better than the accuracy of similar applications we've built with a combination of methods Self-Organizing Map and KNearest Neighbor.

Copyrights © 2012






Journal Info

Abbrev

kursor

Publisher

Subject

Computer Science & IT Decision Sciences, Operations Research & Management

Description

Jurnal Ilmiah Kursor is published in January 2005 and has been accreditated by the Directorate General of Higher Education in 2010, 2014, 2019, and until now. Jurnal Ilmiah Kursor seeks to publish original scholarly articles related (but are not limited) to: Computer Science. Computational ...