<p>Given that substantial hydrocarbons remain locked within source rock formations such as shale, unrecoverable by conventional extraction methods, hydraulic fracturing has become a pivotal technology for exploiting these unconventional resources. Although optimal perforation design is essential to improving hydraulic fracturing efficiency, the complex interactions among multiple perforations during the fracturing process led to markedly different outcomes depending on the perforation distribution pattern. This study leverages an integrated framework of discrete element method (DEM) simulations and deep learning (DL) to classify and evaluate perforation layouts for improved fracturing effectiveness. First, a theoretical model for a single perforation, based on the Kirsch solution from elasticity theory, was developed to elucidate the mechanism by which injection pressure alters the surrounding stress field prior to fracture initiation. Next, a numerical model for hydraulic fracturing was constructed using DEM, wherein the material parameters of the shale were calibrated with machine learning techniques. The consistency observed between the theoretical predictions, numerical results, and experimental data validates the adopted methodology. Further comparative analysis through DEM simulations of single and multi-perforation hydraulic fracturing revealed that reasonable multi-perforation configurations can induce a more favorable stress environment for fracture propagation, promoting the formation of complex fracture networks. On this basis, a multi-perforation DEM model was developed to simulate fracture initiation and propagation from randomly distributed perforations. The simulated dataset was used to train the developed (deep neural networks) DNN classification model for predicting fracturing effects of different multi-perforation configurations. Its performance was validated by confusion matrix analysis and ROC curve evaluation, which demonstrates its high predictive accuracy. The proposed model accurately captures the intrinsic relationship between perforation parameters and fracture morphology, thereby elucidating the underlying mechanisms. The fracture–perforation interaction analysis framework established in this study provides a theoretical basis and practical guidance for optimizing perforation design in large-scale hydraulic fracturing operations.</p>