<![CDATA[The rise in obesity cases in various countries, including Indonesia, has become a serious public health problem because it increases the risk of chronic diseases and affects individuals' psychological aspects. One of the main challenges in obesity management is the differences in obesity types in each individual, which are influenced by various factors. Therefore, accurate classification methods are needed to ensure more targeted treatment. In this context, machine learning-based technology is a potential solution for classifying obesity types. However, variations in individual characteristics make the classification process complex, as models often struggle to accurately distinguish obesity types. To overcome this problem, the Decision Tree algorithm was chosen because of its easy-to-interpret results. However, using Decision Tree with default parameters on datasets with many attributes and high variation tends to cause overfitting and decrease accuracy. Furthermore, Decision Tree performance is highly dependent on hyperparameter settings, requiring optimization techniques to achieve optimal results. Based on this, this study aims to optimize the Decision Tree algorithm using GridSearchCV to obtain the most optimal parameters to improve model performance in obesity type classification. The dataset used is from the UCI Machine Learning Repository, consisting of 2,111 rows of data and 17 attributes. Based on the initial test results, the default model achieved 92.58% accuracy, 92.58% recall, 92.66% precision, and 92.56% F1-score. After optimization, the accuracy increased to 95.69%, 95.69% recall, 95.72% precision, and 95.67% F1-score. The 3.1% increase in accuracy demonstrates the effectiveness of GridSearchCV in improving Decision Tree performance, resulting in a more accurate and stable prediction model. This research is expected to contribute as a basis for decision-making in early detection and prevention and treatment of obesity more efficiently and effectively.]]>
Copyrights © 2025
