Machine Learning-Based Daily Rainfall Data Gap-Filling Method: A Case Study of DKI Jakarta Area

Rain is a hydrometeorological event that cannot be separated from human life. Rainfall is an important part of the process of human life ranging from agriculture, marine, to the environment and natural disasters. Some rainfall measurement equipment was invented to help measure rainfall intensity. However, problems occur in the density of the rainfall data. Many factors cause rainfall data vacuum. This study aims to overcome the problem of rainfall data void by conducting gap-filling based on machine learning (ML). A total of seven rain gauge locations in DKI Jakarta area were used to take daily rainfall data as samples. Some ML used for gap-filling in this study such as extreme gradient boosting (XGB), k-nearest neighbors (KNN), random forest (RF), stochastic gradient descent (SGD), and multi-layer perceptron (MLP). The results of the study obtained were that KNN modeling dominated for the best performance value in training data, but in SGD testing data dominated for the best performance with the average value of RMSE, MAE, and R
²
was 5.95, 2.40, and 0.87, respectively.