Methods, systems, and computer program products for category oversampling for imbalanced machine learning are provided herein. This permits bonding to be finished, corresponding to each wire bonding section at all times, in a shortest time without need of application of unnecessary ultrasonic vibration and pressing force after completion of welding in spite of dispersion of the welding characteristic of the individual bonding pads. As a result, the percent transformation into cubic boron nitride can be considerably enhanced, and the cubic boron nitride obtained exhibits high mechanical strength. The LDAP server analyzes the priority and dependencies to determine the order of each plug-in or individual callback relative to one another. A related method is also described.