Risk Identification Methods In Software Testing

December 05 10:00 2011 by Nataliia Vasylyna Print This Article

Fault distribution is very uneven for the majority of software, not depending on their size, functionality, implementation language and other features.

Much empirical evidence has accumulated over the years to support the so-called 80:20 principle. It states that 20% of the software elements are answerable for 80% of the troubles.

Such problematic elements may commonly be described by specific estimation properties about their design, size, complexity, change history.

Because of the uneven fault distribution among software elements, there is a huge need for risk identification methods to analyze these estimation data so that inspection, software testing and other quality assurance activities can be concentrated on such potentially high-defect elements.

There are several risk detecting methods:

tree-based modeling
traditional statistical analysis methods
neural networks
learning algorithms
pattern matching methods
principal component and discriminant analysis

These methods can be described by such features as:

exactness
presence of tool support
ease of result interpretation
simplicity
stability
creative info
early presence
manual for quality betterment

Correct risk detecting methods may be picked to fit specific application environments with the goal to detect high-risk software elements for focused inspection and software testing.