Vandi Akhmad Royan, Lilik Anifah, Hapsari Peni Agustin Tjahyaningtijas
Roads are essential infrastructure; therefore, regular maintenance is vital. Classification of road damage is a necessary step in automatic road. Manually classifying road damage can be time-consuming and may introduce uncertainty due to the subjective nature of the officer's evaluation of road conditions. Various machine learning strategies, such as decision trees, KNN, and SVM, are employed to automate the categorization of road damage. The study presents a method that depends on the Random Forest (RF) algorithm for classifying road damage, supported by second-order feature extraction. The selection of this technique for extracting the second-order feature is based on its ability to capture the interrelationships among piksel, provide more detailed information about the image, and provide various textual characteristics that aid in a more comprehensive and extensive analysis. The study also compared the accuracy of various machine-learning systems. The experiment findings demonstrated that the Random Forest method exceeded other strategies by achieving an accuracy rate of 9 9 %. At the same time, SVM, KNN, and Decision Tree had an accuracy value of 9 6. 3 3 %, 9 8. 3 3 %, and 9 8 %. The Random Forest algorithm's high accuracy is due to its ensemble approach, which enhances generalization and reduces overfitting, rendering it a superior choice for complex classification tasks in real-world applications. © 2024 IEEE.
Universitas Negeri Surabaya, Department of Electrical Engineering, Surabaya, Indonesia
