Investigating the utility of oblique tree-based ensembles for the classification of hyperspectral data

Nitesh Poona, Adriaan van Niekerk, Riyad Ismail

    Research output: Contribution to journalArticle

    5 Citations (Scopus)

    Abstract

    Ensemble classifiers are being widely used for the classification of spectroscopic data. In this regard, the random forest (RF) ensemble has been successfully applied in an array of applications, and has proven to be robust in handling high dimensional data. More recently, several variants of the traditional RF algorithm including rotation forest (rotF) and oblique random forest (oRF) have been applied to classifying high dimensional data. In this study we compare the traditional RF, rotF, and oRF (using three different splitting rules, i.e., ridge regression, partial least squares, and support vector machine) for the classification of healthy and infected Pinus radiata seedlings using high dimensional spectroscopic data. We further test the robustness of these five ensemble classifiers to reduced spectral resolution by spectral resampling (binning) of the original spectral bands. The results showed that the three oblique random forest ensembles outperformed both the traditional RF and rotF ensembles. Additionally, the rotF ensemble proved to be the least robust of the five ensembles tested. Spectral resampling of the original bands provided mixed results. Nevertheless, the results demonstrate that using spectral resampled bands is a promising approach to classifying asymptomatic stress in Pinus radiata seedlings.

    Original languageEnglish
    Article number1918
    Pages (from-to)1-16
    Number of pages16
    JournalSensors (Switzerland)
    Volume16
    Issue number11
    DOIs
    Publication statusPublished - 15 Nov 2016

    Fingerprint

    Classifiers
    Spectral resolution
    Pinus
    Support vector machines
    Seedlings
    spectral bands
    classifiers
    classifying
    Forests
    Least-Squares Analysis
    spectral resolution
    ridges
    regression analysis

    Cite this

    Poona, Nitesh ; van Niekerk, Adriaan ; Ismail, Riyad. / Investigating the utility of oblique tree-based ensembles for the classification of hyperspectral data. In: Sensors (Switzerland). 2016 ; Vol. 16, No. 11. pp. 1-16.
    @article{b3c2ef5974df4e43bce64e3f62276c22,
    title = "Investigating the utility of oblique tree-based ensembles for the classification of hyperspectral data",
    abstract = "Ensemble classifiers are being widely used for the classification of spectroscopic data. In this regard, the random forest (RF) ensemble has been successfully applied in an array of applications, and has proven to be robust in handling high dimensional data. More recently, several variants of the traditional RF algorithm including rotation forest (rotF) and oblique random forest (oRF) have been applied to classifying high dimensional data. In this study we compare the traditional RF, rotF, and oRF (using three different splitting rules, i.e., ridge regression, partial least squares, and support vector machine) for the classification of healthy and infected Pinus radiata seedlings using high dimensional spectroscopic data. We further test the robustness of these five ensemble classifiers to reduced spectral resolution by spectral resampling (binning) of the original spectral bands. The results showed that the three oblique random forest ensembles outperformed both the traditional RF and rotF ensembles. Additionally, the rotF ensemble proved to be the least robust of the five ensembles tested. Spectral resampling of the original bands provided mixed results. Nevertheless, the results demonstrate that using spectral resampled bands is a promising approach to classifying asymptomatic stress in Pinus radiata seedlings.",
    keywords = "Hyperspectral data, Oblique tree-based ensembles, Pinus radiata, Spectral resampling",
    author = "Nitesh Poona and {van Niekerk}, Adriaan and Riyad Ismail",
    year = "2016",
    month = "11",
    day = "15",
    doi = "10.3390/s16111918",
    language = "English",
    volume = "16",
    pages = "1--16",
    journal = "Sensors (Switzerland)",
    issn = "1424-8220",
    publisher = "MDPI AG",
    number = "11",

    }

    Investigating the utility of oblique tree-based ensembles for the classification of hyperspectral data. / Poona, Nitesh; van Niekerk, Adriaan; Ismail, Riyad.

    In: Sensors (Switzerland), Vol. 16, No. 11, 1918, 15.11.2016, p. 1-16.

    Research output: Contribution to journalArticle

    TY - JOUR

    T1 - Investigating the utility of oblique tree-based ensembles for the classification of hyperspectral data

    AU - Poona, Nitesh

    AU - van Niekerk, Adriaan

    AU - Ismail, Riyad

    PY - 2016/11/15

    Y1 - 2016/11/15

    N2 - Ensemble classifiers are being widely used for the classification of spectroscopic data. In this regard, the random forest (RF) ensemble has been successfully applied in an array of applications, and has proven to be robust in handling high dimensional data. More recently, several variants of the traditional RF algorithm including rotation forest (rotF) and oblique random forest (oRF) have been applied to classifying high dimensional data. In this study we compare the traditional RF, rotF, and oRF (using three different splitting rules, i.e., ridge regression, partial least squares, and support vector machine) for the classification of healthy and infected Pinus radiata seedlings using high dimensional spectroscopic data. We further test the robustness of these five ensemble classifiers to reduced spectral resolution by spectral resampling (binning) of the original spectral bands. The results showed that the three oblique random forest ensembles outperformed both the traditional RF and rotF ensembles. Additionally, the rotF ensemble proved to be the least robust of the five ensembles tested. Spectral resampling of the original bands provided mixed results. Nevertheless, the results demonstrate that using spectral resampled bands is a promising approach to classifying asymptomatic stress in Pinus radiata seedlings.

    AB - Ensemble classifiers are being widely used for the classification of spectroscopic data. In this regard, the random forest (RF) ensemble has been successfully applied in an array of applications, and has proven to be robust in handling high dimensional data. More recently, several variants of the traditional RF algorithm including rotation forest (rotF) and oblique random forest (oRF) have been applied to classifying high dimensional data. In this study we compare the traditional RF, rotF, and oRF (using three different splitting rules, i.e., ridge regression, partial least squares, and support vector machine) for the classification of healthy and infected Pinus radiata seedlings using high dimensional spectroscopic data. We further test the robustness of these five ensemble classifiers to reduced spectral resolution by spectral resampling (binning) of the original spectral bands. The results showed that the three oblique random forest ensembles outperformed both the traditional RF and rotF ensembles. Additionally, the rotF ensemble proved to be the least robust of the five ensembles tested. Spectral resampling of the original bands provided mixed results. Nevertheless, the results demonstrate that using spectral resampled bands is a promising approach to classifying asymptomatic stress in Pinus radiata seedlings.

    KW - Hyperspectral data

    KW - Oblique tree-based ensembles

    KW - Pinus radiata

    KW - Spectral resampling

    UR - http://www.scopus.com/inward/record.url?scp=84995807152&partnerID=8YFLogxK

    U2 - 10.3390/s16111918

    DO - 10.3390/s16111918

    M3 - Article

    VL - 16

    SP - 1

    EP - 16

    JO - Sensors (Switzerland)

    JF - Sensors (Switzerland)

    SN - 1424-8220

    IS - 11

    M1 - 1918

    ER -