Klasifikasi Suhu Pada Peralatan Listrik Tegangan Tinggi Menggunakan K- Nearest Neighbor
Abstrak
The application of artificial intelligence (AI) in classifying safe or hazardous conditions based on thermal images is an innovation that has not been widely developed. Thermal images represent the temperature of an object, where white indicates high temperatures that may be hazardous, while darker colors signify lower, relatively safe temperatures. In this study, a feature extraction approach was used to obtain the percentage of white and non-white colors as classification parameters. The research process began by using 10 thermal images as training data and 2 thermal images as test data. The next step was calculating the distance between each test data and all training data using the Euclidean Distance method. The distance calculation results were then sorted based on proximity, followed by classification using different values of K (the number of nearest neighbors), ranging from K = 1 to K = 7. The test results showed that this method achieved a classification accuracy of 71.42%. This accuracy is still relatively low due to the limited number of training data in the hazardous category, making it difficult for the system to recognize patterns effectively. Additionally, when K values exceed 3, classification accuracy tends to decrease because too many neighbors are considered. To improve accuracy, it is necessary to optimize the selection of K values, increase the amount of training data, and apply data preprocessing techniques to enhance the quality of extracted features. This way, the system can become more reliable and accurate in classifying safe or hazardous conditions based on thermal images.
Identifier
- DOI: https://doi.org/10.24912/tesla.v27i1.33225
- OpenAlex: https://openalex.org/W4416847844

