Ike Fitriyaningsih, Mukhammad Solikhin, Junita Amalia
The water level in a river is an alarm that floods will occur. If a river overflows, then the area around the river will flood. Flooding is one of the most harmful natural disasters in terms of both humanity and economics. Flooding occurs when large amounts of water flood the surface of the soil which is usually dry. Predictions are needed to prevent the impact of events by providing information on the potential for flooding in the area. Deli Serdang Regency is an area that surrounds Medan City as the center of North Sumatra Province which often floods because it is the longest and largest river basin known as the Ular River. Backpropagation Neural Network (BPNN) and Adaptive Neuro Fuzzy Inference System (ANFIS) are machine learning methods that is expected to be able to predict the water level of Ular river so that it can be an early warning when floods occur. Data was used to build model is five years water level daily data from 2014 until 2018. The experiment was carried out 3 times with the dataset divided 90:10, 80:20 and 70:30. The BPNN and ANFIS models were compared using the Root Mean Square Error (RMSE) criteria and the best model obtained was used to predict water levels. Based on the results of experiments carried out on the BPNN and ANFIS methods the smallest RMSE value obtained was in the experiment using a 70:30 dataset division. The RMSE value for the BPNN method is 0.00080140 and the RMSE value for the ANFIS method is 0.00080066. © 2025 American Institute of Physics Inc.. All rights reserved.
Study Program of Artificial Intelligence, Universitas Negeri Surabaya, Surabaya, Indonesia; Study Program of Mathematics, Universitas Negeri Malang, Malang, Indonesia; Study Program of Information Systems, Institut Teknologi Del, Toba, Indonesia
