Implementation of an Adaptive Neuro-Fuzzy Inference System with Particle Swarm Optimization (ANFIS-PSO) for Rainfall Prediction in Sumatera Institute of Technology (ITERA)

Rainfall is one of the weather parameters that have an important role in daily life. However, the occurrence of rainfall cannot be determined with certainty. Therefore, it is necessary to do forecasting or prediction to estimate how much rainfall intensity will occur using past time-series data. In the development forecast phase, one of them is that it combines Adaptive Neuro-Fuzzy Inference System and Particle Swarm Optimization (ANFIS-PSO) to produce better results in the form of optimal prediction values. There are eight model scenarios with two trial pattern treatments that will be tested in this study. The first trial has a pattern of independent input variables (temperature, humidity, solar radiation, wind direction, wind speed, and pressure) and rainfall output variables. The second trial has a pattern of independent input variables plus rainfall data, and the output variable is rainfall on the next day. The test results concluded that the seventh model scenario with the first trial using wind direction, humidity, temperature, solar radiation, and wind speed data as input variables and rainfall as the output variable produced the lowest RMSE value of 0.0237; MSE value of 0.0006; and the correlation value of 0.3899.