The prediction of thunderstorm activity is not only significant for weather forecasting but also for the standardization of risk assessment as in the aviation industry or emergency unit purposes. This paper aimed to develop a prediction of thunderstorm occurrences using a nonlinear model. For this work, the data used for a case study is one-year (1 January 2012 to 31 December 2012) located in a tropical area. The Jacobi algorithm has been employed to construct a prediction model with six combinations of input and one output (target). The predicted target is thunderstorm occurrence. The parameter input is surface pressure, air temperature, relative humidity, clouds, precipitable water vapor, and precipitation. The result obtained a better fit prediction model with four optimum parameters and estimation errors of 5.73%. May and October are the highest occurrences of thunderstorms where prediction errors were found high during the intermonsoon season.