PREDICTION OF CLOUD TOP TEMPERATURE THROUGH THE IMPLEMENTATION OF THE EXTREME LEARNING MACHINE ALGORITH TO IDENTIFY POTENTIAL HAIL

• Main Author: Aulya Rasyada, Annisa
• Format: Final Project
• Language: Indonesia
• Online Access: https://digilib.itb.ac.id/gdl/view/63742
• Institution: Institut Teknologi Bandung

Extreme weather phenomena such as hail that occur in a short time require fast data analysis with a short data period. A renewable method of making short-term predictions is by utilizing machine learning. Extreme Learning Machine (ELM) method is part of machine learning has a good and fast learning rate with a variety of data compared to other conventional methods. Through this research, ELM was tested through a learning process with sampling resolution per 10 minutes. Himawari-8 data processing with Cloud Top Temperature (CTT) and Rainfall Rate (rfr) parameters in June 2021 when hail occurred in the Majalaya region of West Jav was carried out using python. The prediction process is carried out using predictant data of 1 grid, 5 grids, and 9 grids around the AWS point which is used as the center of the hail prediction domain that occurred at that time. Data that has been divided into 2 parts, namely training data and testing data, obtains test results on the testing data, namely the average accuracy value is 50% with a minimum RMSE value of 0.1726 at a training speed of 12.605 seconds against the 2, 4 grid area, 5,6,8 using a lag time of 10 minute. While the predictions made for the case of hail events in the Majalaya region gave the most optimum RMSE value of 0.121 at a lag time of 10 minutes on the 5th grid area. Implementation using the ELM algorithm can be developed by upgrading the data and parameters to be used.