In order to construct a prediction model with high prediction accuracy and strong generalization ability for hydrothermal biofuel hydrogen-rich content, and to deeply explore the laws and mechanisms of hydrothermal conversion of biomass, this paper, based on 243 groups of experimental data collected in the literature for hydrothermal preparation of hydrogen-rich biofuel from high-moisture, high-fat wastes such as diseased and dead livestock and poultry, algae, etc., and adopting two types of high-fitness machine learning algorithms, namely, Random Forest and Extreme Gradient Enhancement Tree, establishes a high-precision and wide-area hydrothermal Based on the mathematical prediction model, the contribution of hydrothermal oil formation conditions, the local dependence response behavior and its mutual coupling law were analyzed by using SHAP interpretable technology and local dependence analysis method. The results showed that: the lipid content and hydrogen content in the high-moisture, high-fat waste were the determining factors for the preparation of hydrogen-rich biofuel, and they ranked the top two in terms of their contribution to the enrichment of hydrogen in the oil phase, which could significantly affect the accumulation of hydrogen in the oil phase; with the increase in the hydrogen content of the feedstock, the hydrogen content in the oil phase was enhanced, indicating that hydrogen-rich feedstock provided a convenient condition for the preparation of hydrogen-rich biofuel, with an enhancement effect of up to 4 wt%. And the higher heating value of the feedstock was the main inhibiting factor for the hydrogen-enriched behavior of biofuel, with an inhibiting effect of up to 4 wt%. In addition, from the interactions among the characterization categories, the coupling between elemental information, industrial information and biomass component information is strong, but the local coupling between feedstock properties and operating conditions is small.