Prediksi Banjir Di Dki Jakarta Dengan Menggunakan Algoritma K-Means Dan Random Forest

Abstract

This research aims to develop a flood prediction method that can be used to implement effective prevention and mitigation measures in dealing with frequent natural disasters in DKI Jakarta. The approach used in this study involves the utilization of Machine Learning techniques with a combination of K-Means and Random Forest algorithms. Historical data on water gates, water levels, and other relevant factors are used as inputs for the development of the flood prediction model. The K-Means method is employed to cluster the water level data, and the results of the K-Means clustering process are then used as parameters in the Random Forest method. A total of 20 experiments were conducted, varying the value of k from 1 to 20 in the K-Means algorithm. The experimental results show that the best accuracy and f-1 score were achieved at k=14, with an accuracy rate of 95% and an f-1 score of 90%. This indicates that the developed flood prediction model is capable of providing accurate and reliable predictions in identifying flood potential. This research holds significant implications for flood management in vulnerable cities. With an effective flood prediction method, prevention and mitigation measures can be implemented more efficiently, thereby reducing the negative impacts caused by floods.