Artificial Neural Network for Rainfall Prediction Base on Historical Rainfall Data by Day

Abstract

Increasingly erratic rainfall has a significant effect on agriculture, and health. One field of agriculture that utilizes rainfall patterns is paddy fields. Rice farmers need a calculation of rainfall to determine the process of hatching and planting rice, especially for farmers who have rainfed rice fields. While in the health sector, rainfall greatly determines the growth of mosquitoes which can lead to dengue fever. The health office needs calculations to anticipate the number of people affected by dengue fever. In this study two data modeling were made in predicting rainfall in the Banyumas area, one data model or scenario using weather data other than rainfall such as temperature, humidity, wind speed, duration of solar radiation. While scenario two only uses historical rainfall data per day. In making models and testing data using the Artificial Neural Network (ANN) algorithm. The results showed that scenario two was better than scenario one after being evaluated using the Root Mean Square Error (RMSE). The RMSE obtained in scenarios one and two are 21,667 and 20,448, respectively. It can be concluded that to predict rainfall in the Banyumas area it is better to use historical rainfall data per day compared to using other weather data.

Forecast Rainfall Density by Utilizing Machine Learning Models

“…Since a significant amount of time ago, researcher have relied on three primary types of models to make accurate precipitation forecasts: linear, statistical, and conceptual models. In previous research, many researchers [6]- [8] have used Artificial neural network (ANN) for forecast rainfall. [8] proposed an artificial neural network model to predict rainfall for Columbia summer season.…”