Two approaches in Empirical Software Engineering

1
Two approaches in Empirical Software Engineering

• Kenichi Matsumoto
• Nara Institute of Science and Technology

2
Two approaches

• Goal-driven Approach (GDA)
• Goal is prescribed and empirical data is
  collected to achieve such specific goal.
• Model to clarify relationship between the goal
  and data is needed.
• GQM Model, ISO Measurement Information Model
• Data-driven Approach (DDA)
• By using large amount of stored data, the
  relationship among software product, process, and
  resource is revealed.
• Visualization and data mining techniques are
  needed.
• Project Visualization, Association Analysis

3
Goal/Question/Metric (GQM) Model
Measurement Goal
Question to evaluate Goal
Data set to answer Question
V. R. Basili and D. M. Weiss A methodology for
collecting valid software engineering data, IEEE
Transactions on Software Engineering, Vol.SE-10,
No.6, pp.728-738, 1984.
4
Association Analysis

• Useful method for discovering interesting
  relationships hidden in large data sets.
• The uncovered relationships can be represented in
  the form of association rules X -gt Y
• 60 of bug reports have Change Request class
  -gt the requirements are unstable
• The strength of an association rule can be
  measured in terms of its support and confidence.
• Support how often a rule is applicable to a
  given data set.
• Confidence Conditional probability P(YX).

5
Integration of Two Approaches
Threshold, Rule, Scale,
Data-driven
Integrated Empirical Approach in SE
AssociationAnalysis
Visualization
Multi-DimensionalScaling
AnalysisModel
Case-BasedReasoning
ISO MI Model
Conventionalmetrics
GQM Model
New metrics
Goal-driven
6
GQM with Association Rules

• Model (Question)
• if ( FCM gt 1.0 and (LCC/LOC gt 5) and (60 of bug
  reports have Change Request class ) then the
  requirements are unstable
• Metric
• FCM Cumulative number of changed files / total
  number of files currently in the
  system.
• LCC the number of lines of code changed.
• LOC the number of lines of code.

Replace them with "Association Rules discovered
in data sets collected in your project or
organization