Differential Analysis and Prediction Based on Independent T-Test and Linear Regression-Stacking Ensemble Algorithm

This paper presents a comprehensive analysis of the differences between experimental and theoretically calculated yield values, employing the Independent T-test to assess statistical significance. By examining the resulting p-values, we determine whether significant discrepancies exist between experimental and theoretical yields. Visual representations of these comparisons are also provided to facilitate subgroup analysis and pinpoint specific mixing ratios where differences are most evident. Furthermore, we develop a predictive model for product yields using a hybrid approach that combines linear regression with a stacking ensemble algorithm. This integrated model, which leverages the strengths of both methodologies through weighted fusion, demonstrates improved predictive performance based on the experimental data.