<p>Double-margin drills are expected to provide radial stability and Ultrasonic-Assisted Drilling (UAD) can reduce cutting forces and torque, but the academic literature on double-margin drills is limited, and their combined effect with UAD is not well understood. To investigate this interactive effect, this paper analyzes the geometric accuracy of holes in Aluminum 7075-T6 using a full factorial design of experiments. The influence of drill margin, ultrasonic-assistance, spindle speed, and feed rate on hole cylindricity and circularity errors at entry and exit were analyzed. Analysis of Variance is used to identify which factors and interactions are statistically significant. The most significant factor affecting the hole cylindricity error was a strong interaction between the drill margin and the spindle speed. The use of double-margin drills improved the cylindricity by 44% at low speed compared with using single-margin drills. However, the use of double-margin drill degraded the cylindricity by 29% at high speed. For the hole circularity, single-margin drills consistently provided a good performance, while applying UAD was often detrimental particularly when combined with double-margin drills. Furthermore, the exit circularity errors were approximately double those at the entry, highlighting the significant instability of the drill at breakthrough phase. The results revealed complex and often conflicting interactions. To model and accurately predict the conflicting interaction and non-linear behavior, a series of Adaptive Neuro-Fuzzy Inference System models were developed and validated. The central conclusion is that a universal optimization is inadequate; this paper provides both the statistical proof and the predictive models necessary for a criteria-based approach, where parameters are selected based on the single most critical geometric feature.</p>