J.M. Beaver and D. Linton (USA)
Software Metrics, Software Engineering, Design Metrics
Error-prone software modules are costly to a development project because their tendency to err is not discovered until, at best, the Testing phase of the life cycle when correcting faults is expensive. This paper presents techniques for using design metrics to: a. determine whether or not previously tested sets of modules are consistent and b. predict those metric-value combinations in a new, untested set of modules that are most likely to be associated with error-prone modules. Two numerical examples are shown to illustrate the predictive approach and to indicate when the historical data is inconsistent. Thus, the application of these techniques determines that historical data is inconsistent (and, hence, should not be used for predictive purposes) or uses consistent historical data to predict/identify error prone modules for a new set of modules during the design phase. The ability to identify error-prone modules during the design phase provides an early opportunity for corrective action, and thereby saves the cost of redesigning and re-implementing an error-prone module later in the life cycle.
Important Links:
Go Back
