4 research outputs found
rifqiharrys/sdb_gui: SDB GUI 3.4.1
No full text<p>This update comes with an ability to generate a scatter plot of the result.</p>
<p>However, the executable version could not generate the plot without getting the GUI terminated itself. I found that the problem comes from generating executable file using <code>auto-py-to-exe</code> with <code>matplotlib</code> inside the source code. Unfortunately, the problem has not been solved yet and if you want to generate a scatter plot of the result, please use the source code instead of the executable file.</p>
rifqiharrys/sdb_gui: SDB GUI 3.6.1
No full text<h2>Release Note</h2>
<ul>
<li>Remove <code>Auto Negative Sign</code> option and replace it with depth direction assignment on main window</li>
<li>Add depth direction option on data save window</li>
</ul>
rifqiharrys/sdb_gui: SDB GUI 3.5.1
No full text<h2>Release Note</h2>
<ul>
<li>Add train data selection based on attribute selection in addition to random selection (based on percentage)</li>
<li>Add new field containing depth values of the output DEM to its corresponding points in test data file</li>
<li>Add report containing RMSE, MAE, and R squared calculation using generated depth from train data and real depth value from test data (result accuracy) in addition to calculation using generated depth from test data and real depth value from test data (model performance)</li>
<li>Saving scatter plot using executable file now works (surprisingly)</li>
</ul>
COMPARISON OF K-NEAREST NEIGHBOR, MULTIPLE LINEAR REGRESSION, AND RANDOM FOREST CLASSIFIERS FOR DEPTH EXTRACTION IN THE SHALLOW WATER OF KEPULAUAN SERIBU, INDONESIA
No full textSatellite-derived bathymetry is a method used to overcome the limitation of survey vessels when acquiring depth data in shallow waters of less than 2 m, especially depths of 0-2 m. Currently, the SDB method has been widely used to provide shallow water bathymetric data. Besides this method can provide wide coverage of depth data, the availability of multitemporal and multiresolution images allows the method to be categorized as a relatively low-cost method compared to conventional surveys. This study compares SDB methods in deriving depth data using various machine learning algorithms using Sentinel-2A images in Kepulauan Seribu, Indonesia. Three machine learning algorithms were compared, namely K-Nearest Neighbors (KNN), Multiple Linear Regression (MLR), and Random Forest (RF), to observe the best-performing method. SDB was applied by combining echo-sounding measurements and the reflectance of blue, green, red, and near-infrared bands of Sentinel 2A. Our research revealed that RF provided the best accuracy compared to MLR and KNN. However, the resulted depth range could not cover very shallow water depth at 0 m. Only the MLR could detect zero depth, but it has the worst RMSE value. KNN provided a feasible result with slightly higher RMSE compared to RF, nonetheless, it took longer runtime for about 30% higher than RF
