Pemeriksa Jawaban Tulisan Tangan untuk Ujian Pilihan Ganda Menggunakan Hybrid Extreme Learning Convolutional Neural Network Machine

Abstract

In Indonesia, exams can be carried out in various ways depending on the type of implementation, namely in the form of Paper Based Test (PBT), Oral Based Test (OBT), and Computer Based Test (CBT). The type most often used in schools is PBT, which is in the form of essay and multiple choice answers. However, the case is different with the multiple choice exam type. This type of exam is usually used during student graduation exams or better known as the National Examination (UN). In its implementation, the UN applies PBT with the concept of multiple choice questions. PBT applied to the UN uses the Object Character Recognition (OCR) method. However, as time goes by, evaluation of this method occurs. Currently, the PBT exam type is starting to be abandoned and switched to the CBT exam type. However, these two types have their respective advantages and disadvantages. Seeing this opportunity, this research proposes a new solution by combining the weaknesses and strengths of the two types. The solution provided is to utilize artificial intelligence such as OCR by proposing a new method, namely the Hybrid Extreme Convolutional Neural Network Machine.