Monitoring Forest Health Using Hyperspectral Imagery: Does Feature Selection Improve the Performance of Machine-Learning Techniques?
This study analyzed highly correlated, feature-rich datasets from hyperspectral remote sensing data using multiple statistical and machine-learning methods. The effect of filter-based feature selection methods on predictive performance was compared. In addition, the effect of multiple expert-based a...
| Main Authors: | , , , , , |
|---|---|
| Format: | Article |
| Language: | English |
| Published: |
MDPI AG
2021-11-01
|
| Series: | Remote Sensing |
| Subjects: | |
| Online Access: | https://www.mdpi.com/2072-4292/13/23/4832 |
| _version_ | 1797507281383849984 |
|---|---|
| author | Patrick Schratz Jannes Muenchow Eugenia Iturritxa José Cortés Bernd Bischl Alexander Brenning |
| author_facet | Patrick Schratz Jannes Muenchow Eugenia Iturritxa José Cortés Bernd Bischl Alexander Brenning |
| author_sort | Patrick Schratz |
| collection | DOAJ |
| description | This study analyzed highly correlated, feature-rich datasets from hyperspectral remote sensing data using multiple statistical and machine-learning methods. The effect of filter-based feature selection methods on predictive performance was compared. In addition, the effect of multiple expert-based and data-driven feature sets, derived from the reflectance data, was investigated. Defoliation of trees (%), derived from in situ measurements from fall 2016, was modeled as a function of reflectance. Variable importance was assessed using permutation-based feature importance. Overall, the support vector machine (SVM) outperformed other algorithms, such as random forest (RF), extreme gradient boosting (XGBoost), and lasso (L1) and ridge (L2) regressions by at least three percentage points. The combination of certain feature sets showed small increases in predictive performance, while no substantial differences between individual feature sets were observed. For some combinations of learners and feature sets, filter methods achieved better predictive performances than using no feature selection. Ensemble filters did not have a substantial impact on performance. The most important features were located around the red edge. Additional features in the near-infrared region (800–1000 nm) were also essential to achieve the overall best performances. Filter methods have the potential to be helpful in high-dimensional situations and are able to improve the interpretation of feature effects in fitted models, which is an essential constraint in environmental modeling studies. Nevertheless, more training data and replication in similar benchmarking studies are needed to be able to generalize the results. |
| first_indexed | 2024-03-10T04:46:20Z |
| format | Article |
| id | doaj.art-50d9c3d870724ecab7501cd9a214b24b |
| institution | Directory Open Access Journal |
| issn | 2072-4292 |
| language | English |
| last_indexed | 2024-03-10T04:46:20Z |
| publishDate | 2021-11-01 |
| publisher | MDPI AG |
| record_format | Article |
| series | Remote Sensing |
| spelling | doaj.art-50d9c3d870724ecab7501cd9a214b24b2023-11-23T02:57:03ZengMDPI AGRemote Sensing2072-42922021-11-011323483210.3390/rs13234832Monitoring Forest Health Using Hyperspectral Imagery: Does Feature Selection Improve the Performance of Machine-Learning Techniques?Patrick Schratz0Jannes Muenchow1Eugenia Iturritxa2José Cortés3Bernd Bischl4Alexander Brenning5GIScience Group, Department of Geography, Friedrich Schiller University Jena, Loebdergraben 32, 07743 Jena, GermanyGIScience Group, Department of Geography, Friedrich Schiller University Jena, Loebdergraben 32, 07743 Jena, GermanyNEIKER Tecnalia, 48160 Tecnalia, SpainGIScience Group, Department of Geography, Friedrich Schiller University Jena, Loebdergraben 32, 07743 Jena, GermanyDepartment of Statistics, Ludwig-Maximilians-Universität München, Akademiestrasse 1/I, 80799 Munich, GermanyGIScience Group, Department of Geography, Friedrich Schiller University Jena, Loebdergraben 32, 07743 Jena, GermanyThis study analyzed highly correlated, feature-rich datasets from hyperspectral remote sensing data using multiple statistical and machine-learning methods. The effect of filter-based feature selection methods on predictive performance was compared. In addition, the effect of multiple expert-based and data-driven feature sets, derived from the reflectance data, was investigated. Defoliation of trees (%), derived from in situ measurements from fall 2016, was modeled as a function of reflectance. Variable importance was assessed using permutation-based feature importance. Overall, the support vector machine (SVM) outperformed other algorithms, such as random forest (RF), extreme gradient boosting (XGBoost), and lasso (L1) and ridge (L2) regressions by at least three percentage points. The combination of certain feature sets showed small increases in predictive performance, while no substantial differences between individual feature sets were observed. For some combinations of learners and feature sets, filter methods achieved better predictive performances than using no feature selection. Ensemble filters did not have a substantial impact on performance. The most important features were located around the red edge. Additional features in the near-infrared region (800–1000 nm) were also essential to achieve the overall best performances. Filter methods have the potential to be helpful in high-dimensional situations and are able to improve the interpretation of feature effects in fitted models, which is an essential constraint in environmental modeling studies. Nevertheless, more training data and replication in similar benchmarking studies are needed to be able to generalize the results.https://www.mdpi.com/2072-4292/13/23/4832hyperspectral imageryforest health monitoringmachine learningfeature selectionmodel comparison |
| spellingShingle | Patrick Schratz Jannes Muenchow Eugenia Iturritxa José Cortés Bernd Bischl Alexander Brenning Monitoring Forest Health Using Hyperspectral Imagery: Does Feature Selection Improve the Performance of Machine-Learning Techniques? Remote Sensing hyperspectral imagery forest health monitoring machine learning feature selection model comparison |
| title | Monitoring Forest Health Using Hyperspectral Imagery: Does Feature Selection Improve the Performance of Machine-Learning Techniques? |
| title_full | Monitoring Forest Health Using Hyperspectral Imagery: Does Feature Selection Improve the Performance of Machine-Learning Techniques? |
| title_fullStr | Monitoring Forest Health Using Hyperspectral Imagery: Does Feature Selection Improve the Performance of Machine-Learning Techniques? |
| title_full_unstemmed | Monitoring Forest Health Using Hyperspectral Imagery: Does Feature Selection Improve the Performance of Machine-Learning Techniques? |
| title_short | Monitoring Forest Health Using Hyperspectral Imagery: Does Feature Selection Improve the Performance of Machine-Learning Techniques? |
| title_sort | monitoring forest health using hyperspectral imagery does feature selection improve the performance of machine learning techniques |
| topic | hyperspectral imagery forest health monitoring machine learning feature selection model comparison |
| url | https://www.mdpi.com/2072-4292/13/23/4832 |
| work_keys_str_mv | AT patrickschratz monitoringforesthealthusinghyperspectralimagerydoesfeatureselectionimprovetheperformanceofmachinelearningtechniques AT jannesmuenchow monitoringforesthealthusinghyperspectralimagerydoesfeatureselectionimprovetheperformanceofmachinelearningtechniques AT eugeniaiturritxa monitoringforesthealthusinghyperspectralimagerydoesfeatureselectionimprovetheperformanceofmachinelearningtechniques AT josecortes monitoringforesthealthusinghyperspectralimagerydoesfeatureselectionimprovetheperformanceofmachinelearningtechniques AT berndbischl monitoringforesthealthusinghyperspectralimagerydoesfeatureselectionimprovetheperformanceofmachinelearningtechniques AT alexanderbrenning monitoringforesthealthusinghyperspectralimagerydoesfeatureselectionimprovetheperformanceofmachinelearningtechniques |