Software fault count and density prediction

Software fault count and density prediction Dr. Bob Hughes, Franklyn Young-Martos, Alan Cunliffe Synopsis It has often been found that the majority of faults in a software application are caused by a small number of software components. Previous work on fault distribution has also tended to show that smaller modules often have larger fault densities than ones with more lines of code. Work carried out at Ericsson, Burgess Hill, is described which analysed the fault counts and densities in both newly developed and subsequently modified software modules. These analyses studied the relationship between fault counts and densities and other measurable characteristics of the components. On the basis of this work, possible prediction models for fault counts are suggested.