<p>Aiming at the problem of insufficient effective tooth width in the debugging and production of a spur gear, a small-sample data-driven “experimental design-model construction-algorithm optimization-decision verification” multi-objective optimization and decision-making method for spur gear cold extrusion forming process parameters was proposed. Small-sample data-driven algorithms were introduced to construct a multi-objective optimization model for the relationship between process parameters and the bottom protrusion length, effective tooth width, and maximum effective stress of the tooth profile concave die of the extrusion gear. The improved multi-objective grey wolf algorithm was applied to optimize the multi-objective optimization model and obtain a non-dominated solution set. The subjective–objective comprehensive entropy weight optimal solution distance method was used to evaluate and determine the optimal combination of process parameters. The dual verification results of numerical simulation and physical process experiments show that the simulation results were basically consistent with the process experiments, and the effective tooth width and tool service life met the design requirements. The optimization and decision-making method based on a small-sample data-driven approach proposed in this study breaks through the technical limitations of insufficient accuracy of multi-objective models under small samples and provides an efficient solution path for similar complex process optimization problems.</p>