PENGOLAHAN DATA RAWIN SONDE MENGGUNAKAN METODE MACHINE LEARNING UNTUK IDENTIFIKASI POTENSI PETIR DI WILAYAH CILACAP

Lightning is the release of electric charge in the air due to the difference in the electric field between two masses with different electric charges to reach equilibrium (Griffith, 1995). The impact of a lightning strike is very dangerous for the safety of life, and can cause material losses because it can burn the object it strikes. By using data obtained from rawin sonde, namely observation of the upper air by releasing a weather balloon carrying a transmitter containing a weather parameter sensor, the atmospheric lability index value is obtained which can be used to identify potential lightning. Atmospheric lability data used are showalter index, lifted index, convergence index, and total-total index. This data is commonly used by various meteorological agencies in other countries to predict the potential for bad weather in the future, including lightning. It's just that the index threshold is in fact not the same for all regions, depending on the weather character of the region, especially for tropical areas such as Indonesia which have weather parameter values such as air pressure above sea level which tend to be homogeneous. Given their superiority, the KNN and Naive Bayes algorithms will be used to identify potential lightning for the next six hours in the area observed by the weather observer at the Tunggul Wulung Meteorological Station, Cilacap. After the implementation of machine learning using the selected algorithm, the prediction accuracy value using the KNN method is 95.04%, and 82.87% using the Naive Bayes method.