REKONSTRUKSI DATA KONSENTRASI KLOROFIL-A DI TELUK JAKARTA MENGGUNAKAN METODE MACHINE LEARNING

• Main Author: Zhafara Nirwana, Muhammad
• Format: Final Project
• Language: Indonesia
• Subjects: Ilmu kebumian
• Online Access: https://digilib.itb.ac.id/gdl/view/63811
• Institution: Institut Teknologi Bandung

Jakarta Bay is a water that often experiences phenomena related to chlorophyll-a such as algae blooms, so the presence of chlorophyll-a concentration data is needed. Chlorophyll-a concentration data can be obtained by direct measurement and satellite imagery. There is a lack of satellite imagery because there is data missing due to the satellite's inability to capture the concentration of chlorophyll-a as it is covered by clouds. In this research, machine learning program with backpropagation neural network algorithm is used. This machine learning program was applied to reconstruct the blank data and reconstruct the occurrence of chlorophyll-a concentrations in Jakarta Bay. Reconstruction of blank data was carried out by means of single step prediction from the previous time and obtained an RMSE value of 1.33 mg/m3. After reconstructing the empty data, the reconstructed data is used as input to reconstruct the occurrence of chlorophyll-a concentration data. This program is made with 3 scenarios, the scenarios are scenario 1 (training data = 3 years), scenario 2 (training data = 5 years), and scenario 3 (training data = 10 years). Accurate results were obtained in reconstructing the concentration of chlorophyll-a where the 3 scenarios had a correlation value of r = 0.99. The results show that scenario 3 is the most accurate scenario compared to the other 2 scenarios with RMSE values ranging from 2.8 mg/m3 (scenario 1), 2.11 mg/m3 (scenario 2), and 1.15 mg/m3 (scenario 3). Based on the division of zones (off, middle, and near the coast) it was found that the offshore area had the best performance for the three scenarios with RMSE values for each scenario of 1.74 mg/m3, 1.25 mg/m3, and 0.87 mg/m3. Based on the division of zones (west, middle, and east coast) it is found that the western region has the best performance for scenario 1 with RMSE values of 2.69 mg/m3 while in scenarios 2 and 3, the eastern region has the best performance. the best with an RMSE value of 1.87 mg/m3 and 1.02 mg/m3. The concentration of chlorophyll-a has a maximum value in the West Season and a minimum in the East Season and nearshore areas have a high concentration of chlorophyll-a while offshore have a lower value.