KLASIFIKASI KEPRIBADIAN MELALUI GAMBAR TULISAN TANGAN MENGGUNAKAN METODE CONVOLUTIONAL NEURAL NETWORK

Abstract

One of the reasons why technology is developing rapidly is because of the effectiveness and efficiency of time. The problem found by the researcher is about psychological science that not many people master, especially the science of reading handwriting so that it requires special training and learning. Researchers aim to build and find the most optimal parameter value for handwriting recognition with an image based on deep learning. Researchers use collecting images with Simple Random Sampling technique to be used as a dataset and AlexNet-based CNN (Convolutional Neural Network) method as a method for testing and experimenting with handwriting image recognition. This research produces the most optimal parameter values based on testing of several existing parameters. The purpose of this research is to facilitate other researchers who want to make deep learning applications for image recognition. The source of the dataset used in this research is obtained from scanned images or photos of handwriting of people around as many as 418 images. Tests were carried out with the MatLab application and several parameters were tested, namely input size, epoch, mini batch size and learning rate. At the end of the test, the researcher obtained a recognition accuracy of 94.17 percent and has given very good results and as expected. Researcher concludes that this handwriting recognition deep learning application runs well. Other researchers can make this research as a reference to develop it even better.

Keywords: Personality Classification, CNN, Handwriting, AlexNet, Matlab