Abstract—Software archives contain historical information
about the development process of a software system. Using
Association rule discovery, the development patterns can be
extracted from these archives. This information is useful to
support software modification activities, as indicated to
software developers which modules are usually modified
together during software maintenance or evolution. All
previous works focused on mining associations with classical
interestingness measure, support-confidence, where some
disadvantage existed. The new Interestingness Measure, named
as support-new confidence, was proposed by Liu et al. to
improve the classical method. In this research, we present the
comparison of the efficiency of applying association rule
discovery on software archive using classical model and Liu et
al.'s model. The experiments were conducted on software
archive of KMyMoney software, an open source financial
software project. The results show that the efficiency of the
rules obtained in new model is higher than the rules obtained in
classical model in navigation scenario.
Index Terms—Association rule discovery, measures, version
control system, software archives interestingness.
Sunchai Pitakchonlasup is with the Appsphere Group Co.,Ltd., Bangkok
Thailand. (e-mail:dz.yoez@gmail.com).
Assadaporn Sapsomboon is with Business Software Devlopment
Program, Faculty of Commerce and Accoutancy, Chulalongkorn University,
Bangkok, Thailand. (e-mail:assadaporn@ acc.chula.ac.th).
Cite:Sunchai Pitakchonlasup and Assadaporn Sapsomboon, "A Comparison of the Efficiency of Applying Association Rule Discovery on Software Archive using Support – Confidence Model and Support – New Confidence Model," International Journal of Machine Learning and Computing vol.2, no. 4, pp. 517-520, 2012.