Imbalanced Learning in Land Cover Classification: Improving Minority Classes’ Prediction Accuracy Using the Geometric SMOTE Algorithm

Abstract

The automatic production of land use/land cover maps continues to be a challenging problem, with important impacts on the ability to promote sustainability and good resource management. The ability to build robust automatic classifiers and produce accurate maps can have a significant impact on the way we manage and optimize natural resources. The difficulty in achieving these results comes from many different factors, such as data quality and uncertainty. In this paper, we address the imbalanced learning problem, a common and difficult conundrum in remote sensing that affects the quality of classification results, by proposing Geometric-SMOTE, a novel oversampling method, as a tool for addressing the imbalanced learning problem in remote sensing. Geometric-SMOTE is a sophisticated oversampling algorithm which increases the quality of the instances generated in previous methods, such as the synthetic minority oversampling technique. The performance of Geometric- SMOTE, in the LUCAS (Land Use/Cover Area Frame Survey) dataset, is compared to other oversamplers using a variety of classifiers. The results show that Geometric-SMOTE significantly outperforms all the other oversamplers and improves the robustness of the classifiers. These results indicate that, when using imbalanced datasets, remote sensing researchers should consider the use of these new generation oversamplers to increase the quality of the classification results.

Addressing the Class Imbalance Problem in the Automatic Image Classification of Coastal Litter From Orthophotos Derived From Uas Imagery

“…This happened when testing against classification thresholding, class weighting and against a baseline experiment where the class imbalance was not taken into account. These results are in agreement with the literature when addressing the class imbalance problem: oversampling approaches often are the best performing ones (Johnson et al, 2013;Buda et al, 2018;Douzas et al, 2019), even if in this case a binary classification problem was considered and an imbalance ratio of around 1:100. Nonetheless the precision-recall curve indicates a smaller difference between oversampling and baseline experiment where the latter even had a marginal improvement when it comes to the pr-AUC.…”

“…Data level methods focus on the training data and on class distribution to reduce/eliminate class imbalance. Two main approaches can be distinguished: oversampling (Johnson et al, 2013;Douzas et al, 2019) and undersampling (Leichtle et al, 2017;Buda et al, 2018). These methods use new image samples generation (oversampling) or modify the subset sample selection (undersampling).…”

“…Other algorithmic methods focus directly on the loss function which penalizes a certain type of errors (Dong et al, 2019). Overall, oversampling is recurrently indicated as the best performing approach in systematic studies (Buda et al, 2018) and for different remote sensing applications (Johnson et al, 2013;Douzas et al, 2019;Xia et al, 2019). This work proposes to analyse and address the class imbalance issue in the automatic image classification of marine litter on coastal environment.…”

Different Spectral Domain Transformation for Land Cover Classification Using Convolutional Neural Networks with Multi-Temporal Satellite Imagery

“…To statistically compare model performance with multiple datasets, Demšar [58] and Garcia and Herrera [59] suggested the Wilcoxon paired rank test and Friedman test, which are nonparametric. The Wilcoxon paired signed-rank test was used only when two models were compared [60][61][62]. The Friedman test was used for the multiple model comparisons [63][64][65] since multiple The first and second rows show the rate of occurrence of line and polygon graphs as density using the reference data.…”

SMOTE-Based Weighted Deep Rotation Forest for the Imbalanced Hyperspectral Data Classification

“…Random oversampling is simply copying the sample of the minority class, which easily leads to overfitting [44] and has little effect on improving the classification accuracy of the minority class. The synthetic minority oversampling technique (SMOTE) is a powerful algorithm that was proposed by Chawla [29] and has shown a great deal of success in various applications [45][46][47]. SMOTE will be described in detail in Section 2.1.…”