Use of near-infrared spectroscopy and multivariate approach for estimating silage fermentation quality from freshly harvested maize

The study aimed to evaluate the most predictive traits of fresh maize and appropriate multivariate approach for estimating silage fermentation quality. use near infrared (NIRs) instruments allowed rapid, accurate cheap analysis. Samples plant (n = 822) from hybrids (Class Cultivar) early late classes, were harvested at three maturity stages: early, medium late, in areas (level input field) ‘low’, ‘medium’ ‘high’ soil fertility, along consecutive years. Several algorithms feature selection, regression, classification machine learning, tested. Maize fermentative quality was summarised through a Fermentative Quality Index (FQI). We found as dry matter (DM), starch, acid detergent lignin (ADL), with negative coefficients, or water-soluble carbohydrates (WSC) positive coefficient. FQI significantly (p < 0.0001) affected by year (negatively 2018), level field (positively high level) stage harvest). satisfying results attained using stepwise regression algorithm (R2 0.48), improved introduction fixed effects 0.55) partial least square discriminant analysis (PLS-DA), which assessed Mattew Correlation Coefficient (MCC) validation (MCC 0.57). Concluding, among tested approaches, linear after PLS could be practical help farmers effective management ensiling process plants, even though environmental conditions should considered improve predictions.HIGHLIGHTSThe prediction FQ harvest would allow farmer tune up processThe portable NIRs successfulDM, starch ADL negatively related index