• • Random Forest model achieved a test set R² of 0.9463, outperforming BP neural network, ANFIS, SVM, and KNN, with minimal overfitting risk, enabling reliable prediction of acidogenic performance in chemical-biological sludge fermentation.
• • pH and VSS were identified as the most critical variables driving acidogenic efficiency, with feature importance scores of 1.17 and 0.894, respectively, indicating that process control should prioritize pH adjustment and organic matter stabilization.
• • Aluminum salts (importance 0.657) exhibited a stronger inhibitory effect on acidogenesis than iron salts (importance 0.527), guiding the need for targeted control of aluminum-based coagulants in sludge management.
• • The proposed optimization pathway 'adjust pH, stabilize organic matter, control aluminum salts' provides a data-driven strategy to enhance short-chain fatty acid production, potentially improving resource recovery efficiency in municipal sludge treatment.