Evaluation of Light Gradient Boosted Machine Learning Technique in Large Scale Land Use and Land Cover Classification

Abstract

The ability to rapidly produce accurate land use and land cover maps regularly and consistently has been a growing initiative as they have increasingly become an important tool in the efforts to evaluate, monitor, and conserve Earth’s natural resources. Algorithms for supervised classification of satellite images constitute a necessary tool for the building of these maps and they have made it possible to establish remote sensing as the most reliable means of map generation. In this paper, we compare three machine learning techniques: Random Forest, Support Vector Machines, and Light Gradient Boosted Machine, using a 70/30 training/testing evaluation model. Our research evaluates the accuracy of Light Gradient Boosted Machine models against the more classic and trusted Random Forest and Support Vector Machines when it comes to classifying land use and land cover over large geographic areas. We found that the Light Gradient Booted model is marginally more accurate with a 0.01 and 0.059 increase in the overall accuracy compared to Support Vector and Random Forests, respectively, but also performed around 25% quicker on average.

A Survey of Ensemble Learning: Concepts, Algorithms, Applications, and Prospects

“…Thirdly, it achieves much higher accuracy than most boosting methods, resulting from the introduction of GOSS and EFB techniques. Lastly, the LightGBM algorithm performs well when trained with large datasets, with a faster training time than the XGBoost algorithm [93]. In terms of disadvantages, the LightGBM can overfit small training datasets easily as it performs better with large datasets.…”

Pixel and object-based land cover mapping and change detection from 1986 to 2020 for Hungary using histogram-based gradient boosting classification tree classifier

“…Many studies have focused on land use/land cover classification in different scenarios. McCarty et al compared three different algorithms (random forest, SVM, and light GBM) for large-scale land use mapping (McCarty et al, 2020). In their classification sce-nario, seven classes were targeted, and the light GBM had the highest overall accuracy (65.3%) and random forest had the lowest overall efficiency (59.4%).…”

Spatial distribution of benthic flora and fauna of coastal placentia bay, an ecologically and biologically significant area of the island of newfoundland, atlantic Canada

“…LightGBM advertises faster training speed, low memory usage (Ke et al, 2017;McCarty et al, 2020), and in-depth control over boosting and model learning via tunable parameters. The latter point, though, requires the user to be confident in their understanding of each parameter.…”