Hyperparameter Tuning and Preprocessing Techniques for Optimizing Javanese Scripts Classification

The Javanese script characters, known as “Aksara Jawa,” is a classical script used primarily in Java, Indonesia. It comprises a unique set of characters that are complex in design and rich in cultural significance. The application of CNNs to the recognition of Javanese characters holds great promise. The CNNs can effectively handle the variations in stroke patterns, shapes, and sizes of the characters, which are characteristic of the Javanese scripts. Hyperparameter tuning is a pivotal process in optimizing the CNNs for the classification of Javanese scripts. The unique and complex structure of Javanese characters requires careful selection and adjustment of hyper-parameters to achieve high accuracy and robust performance. They are optimizing Javanese scripts classification through hyperparameter tuning in a deep learning framework. Adjusting parameters such as learning rate, dropout rate, and optimizer settings significantly enhanced the model's performance. The optimized model achieved a training accuracy of 0.99 and a testing accuracy of 0.94, showcasing its ability to generalize well to unseen data. The F1 scores ranged from 0.84 to 0.99, indicating a balanced performance across script characters. These findings demonstrate the critical role of hyperparameter tuning in improving the accuracy and robustness of deep learning models for Javanese scripts classification.