The other two systems, ME+CR and CRF+CR, are the proposed methods which combine an ML classifier and the rule-based error modifier.
According to the definition of R and M, the difference from the accumulated number of tuples for in the dark line to that in the light line means the amount of improvement when the rule-based error modifier is applied.
The proposed decision function determines the threshold value and prevents the application of the rule-based error modifier from the tuples whose difference value is greater than the threshold value.
The third cluster of outcomes, including "part damaged during repair" "system operated unsafely during maintenance," and "wrong equipment or part installed," was associated with knowledge-based and rule-based errors and, to a lesser extent, with violations.
Fatigue, however, was not associated with an increase in the prevalence of violations, rule-based errors, or knowledge-based errors.
These were violations, slips, knowledge-based errors, and rule-based errors. In most cases, particular contributing factors were associated with an increased incidence of particular errors rather than with an overall increase in all forms of error.