Biochar has gained attention for its role in soil improvement and environmental remediation. Predicting its physicochemical properties from various biomass sources is essential for optimising its applications. Existing linear regression models have investigated the effects of biomass type and pyrolysis conditions on biochar physicochemical properties and soil quality. The models are highly specific and lack predictive accuracy with limited biomass classifications. This study develops a multiple linear regression model, integrating Principal Component Analysis (PCA) to reduce dependent variables and enhance predictions of key biochar properties: pH, Cation Exchange Capacity (CEC), and Electrical Conductivity (EC). Three classification approaches—uncategorised (Combination UC), lignocellulosic analysis (Combination LA), and elemental analysis (Combination EA)—were compared. Classification significantly improved prediction accuracy, with Combination EA outperforming Combination LA. Applying PCA to Combination EA (EAPCA) further enhanced model efficiency, achieving high adjusted R² values for pH, EC, and CEC in woody (0.817, 0.537, 0.875), herbaceous (0.795, 0.759, 0.732), and wet biomass (0.76, 0.787, 0.607) categories. The woody biomass case exhibited the strongest predictive performance for CEC. Key parameters identified through PCA included residence time, heating rate, nitrogen, hydrogen, and H/C ratio. The model's RMSE (15.4778) and R² (0.875) indicate strong predictive capability, explaining 87.5 % of the variance in CEC. This study highlights the effectiveness of classification and PCA in improving biochar property predictions.