Computer self-efficacy (CSE) is a person's judgment of his or her ability to use a computer system. We investigated cognitive engagement, prior experience, computer anxiety, and organizational support as determinants of CSE in the use of a multimedia ERP system's training tool. We also examined the impact of CSE on its acceptance. We determined the benefits of a sequential multi-method approach using structural equation modeling and neural network analysis. High reliability predictions of individual CSE were achieved with a sequential multi-method approach. Specifically, we obtained almost 68% perfect CSE group prediction overall, with almost 85% perfect CSE group prediction using fuzzy sets and over 94% accuracy within one group classification. The resulting CSE assessment and classification enables management interventions, such as allocating users to appropriate instruction for more effective training.