Main Controlling Factors of Tight Oil Fracturing Based on Machine Learning and Engineering Analysis

To resolve the significant disparities in fracturing performance of horizontal wells within Block M of the Songliao Basin, this study employs a framework encompassing parameter dimensionality reduction,algorithm-based ranking, mechanistic interpretation, and simulation validation to identify key controlling factors and optimize fracturing designs. Through Pearson correlation analysis, 16 geological and engineering parameters were reduced to 8 core variables. Coupled with Random Forest and SHAP analyses, the influence weights were quantified, followed by validation via engineering mechanism elucidation and fracturing software simulations. Results indicate that the top four primary influencing factors include reservoir properties, coarse proppant proportion, intra-stage diversion count, and guar gum content in fracturing fluid. Effective stimulation of class I reservoirs serves as the foundation: the proposed optimal scheme—incorporating coarse proppant, guar gum-based fracturing fluids, composite temporary plugging techniques, narrow cluster spacing (5–8 m for class I reservoirs), proppant injection intensity of 2.5–3.0 m3/m, and fluid injection intensity of 25–35 m3/m—enables the designed well to attain an initial daily oil production of 10.5 t with fracture coverage enhanced to 75%. The integrated approach combining data-driven and engineering analyses developed herein provides a transferable technical pathway for optimizing fracturing schemes in tight oil reservoirs.