Composition-Centered Prediction of Kenaf Core Saccharification for Next-Generation Bioethanol via Machine Learning

Yitong Niu^*Ying Ying TyeChee Keong LeeMardiana Idayu AhmadCheu Peng Leh^*

• Universiti Sains Malaysia, Minden Heights, Malaysia

Biomass pretreatment outcomes are heterogeneous across routes and severities, and condition-centred empirical models often fail to generalize beyond the settings on which they were trained. This gap limits early-stage decisions about where to focus costly wet-lab effort. To address this, the study examines a composition-centred surrogate that treats the post-pretreatment solid composition-cellulose, hemicellulose, lignin—as the input space and predicts enzymatic glucose yield as the response. Using n=35 kenaf-core samples under a fixed hydrolysis protocol, random-forest models were tuned by six optimizers. Held-out performance clusters tightly (test R2≈0.49-0.55; RMSE 4.42-4.69 GY%), indicating that attainable accuracy is governed more by model capacity and data coverage than by optimizer choice. Feature diagnostics converge on a cellulose-led mechanism (cellulose positive and monotonic; lignin negative; hemicellulose weaker and context-dependent). Iso-yield maps translate these patterns into feasible composition windows that prioritize high-cellulose/low-lignin targets. Given this accuracy band, the surrogate is best positioned for uncertainty-aware screening to prune unproductive regions before targeted design-of-experiments, rather than replacing it.