Comparison between random forest and support vector machine algorithms for LULC classification

Abstract

Nowadays, machine learning (ML) algorithms have been widely chosen for classifying satellite images for mapping Earth's surface. Support Vector Machine (SVM) and Random Forest (RF) stand out among these algorithms with their accurate results in the literature. The aim of this study is to analyze the performances of these algorithms on land use and land cover (LULC) classification, especially wetlands which have significant ecological functions. For this purpose, Sentinel-2 satellite image, which is freely provided by European Space Agency (ESA), was used to monitor not only the open surface water body but also around Marmara Lake. The performance evaluation was made with the increasing number of the training dataset. 3 different training datasets having 10, 15, and 20 areas of interest (AOI) per class, respectively were used for the classification of the satellite images acquired in 2015 and 2020. The most accurate results were obtained from the classification with RF algorithm and 20 AOIs. According to obtained results, the change detection analysis of Marmara Lake was investigated for possible reasons. Whereas the water body and wetland have decreased more than 50% between 2015 and 2020, crop sites have increased approximately 50%.

Comparison of Three Machine Learning Algorithms Using Google Earth Engine for Land Use Land Cover Classification

“…The accuracy assessment revealed the excellence of the classification map ( AVCI et al 2021 ). The accuracy assessment of the machine learning algorithm was performed after the classification of LULC to evaluate the performance of the models.…”

Comparative analysis of random forest and support vector machine for benthic habitat segmentation

“…7,8,9 However RF and SVM algorithms were considered and suggested to be robust algorithms for image classification producing relatively high and consistent classification and regression accuracies. 10,11 RF, first proposed by Leo Breiman in 2001, is an ensemble learning method which utilizes a combination of multiple decision trees to create a stable classification and regression model. 18 This algorithm implements classification by creating multiple decision trees during the training process and combines them through a voting process producing a final prediction model.…”

“…18,19 But despite the different implementation of classification or regression process, these two algorithms have become commonly used due to their flexibility making them capable of handling large and high-dimensional data with non-linear relationships. 18,19 Between the two algorithms, studies such as Sheykhmousa et 10,12,13,14,15,16,17 With regards modeling of benthic habitats using remote sensing data, supervised image classification using machine learning algorithms like SVM and RF has been common algorithms. 17,21 However, while one algorithm can outperform the other such as in the study of Wicaksono et al in 2019, the use of the dataset and other derivatives have been found to be strongly contributing to the algorithms capability to accurately classify benthic features from the image.…”

Evaluation of different machine learning algorithms for LULC classification in heterogeneous landscape by using remote sensing and GIS techniques

“…For Sentinel-2 the overall accuracy of RF, SVM, DT and CART are 94.8%, 91.79%, 90.84% and 89.30% respectively. In general, RF model gives best performance as it is more reliable and less impacted by its feature [27]. During the process of producing predictions, the final outcome is chosen by averaging the various decision trees that made up the ensemble of decision trees.…”