Fiber Quality Prediction Using Nir Spectral Data: Tree-Based Ensemble Learning VS Deep Neural Networks

The growing applications of near infrared (NIR) spectroscopy in wood quality control and monitoring necessitates focusing on data-driven methods to develop predictive models. Despite the advancements analyzing NIR spectral data, literature science engineering has mainly uti- lizedthe classic model development methods, such as principal component analysis (PCA) regression or partial least squares (PLS) regression, with relatively limited studies conducted evaluating machine learning (ML) models, specifically, artificial neural networks (ANNs). This couldpotentially limit performance specifically for some properties, tracheid width that are both time-consuming tomeasure challenging predict using data. study aims enhance prediction accuracy deep tree-based ensemble algorithms a dataset consisting 2018 samples 692 features (NIR spectra wavelengths). Accord- ingly, were fed into multilayer perceptron (MLP), 1 dimensional-convolutional net- works (1D-CNNs), random forest, TreeNet gradient-boosting, extreme gradient-boosting (XGBoost), light (LGBM). It was interest models without applying PCA assess how effective they would perform when with- out employing dimensionality reduction shown machines outper- formed ANNs regardless number (data dimension). Allthe performed better PCA. is concluded could be effectively used characterization utilizing medium-sized dataset.