The purpose of the work is to demonstrate the usefulness and applicability of audit trail data for modeling, classifying and predicting human user decision behavior using Independent Component Analysis. A hybrid behavior-modeling method, based on independent component analysis and sensitivity analysis, is proposed to directly classify decision behavior in terms of predetermined target measures. A comparison with alternative principal component analysis schemes demonstrates the advantages of the proposed method in terms of algorithmic speed, complexity, practical feasibility and data reduction capability. The capability of the system to provide intelligent support to executive decision making and internal control and to assist in several phases of the decision making process is highlighted.