<![CDATA[This study investigates the use of cosine similarity as a distance metric to enhance Javanese script (Hanacaraka) recognition. A dataset of 4,385 images representing 20 classes was processed using Histogram of Oriented Gradients (HOG) for feature extraction and classified using K-Nearest Neighbor (KNN) and Support Vector Machine (SVM) optimized via GridSearchCV. The SVM with cosine similarity achieved an accuracy of 99.84%, outperforming traditional metrics like Euclidean and Minkowski distances, as well as prior methods such as CNN-SVM and LBP-SVM. Cosine similarity's emphasis on angular relationships, rather than magnitude, made it effective for handling high-dimensional and intricate patterns. ANOVA analysis confirmed significant differences among distance metrics, validating cosine similarity’s advantages. This research supports cultural preservation and advances pattern recognition technologies. Future studies should explore integrating cosine similarity with deep learning models like CNNs or Transformers to improve scalability and adapt to broader datasets, including other traditional scripts.]]>
Copyrights © 2025
