Download as PDF, PPTX
Uploaded byICSM 2011
Metrics - Using Source Code Metrics to Predict Change-Prone Java Interfaces
The document discusses a study on the effectiveness of source code metrics, specifically the interface usage cohesion (IUC) metric, in predicting change-prone Java interfaces through analysis of correlation with code changes. It evaluates multiple projects using various predictive models and machine learning algorithms, concluding that IUC shows a stronger correlation with changes than traditional metrics. Additionally, the study highlights implications for researchers, quality engineers, and developers, as well as future research directions.
More Related Content
Similar to Metrics - Using Source Code Metrics to Predict Change-Prone Java Interfaces
Metrics - Using Source Code Metrics to Predict Change-Prone Java Interfaces
- 1. Using Source CodeMetrics to Predict Change- Prone Java Interfaces Daniele Romano and Martin Pinzger Williamsburg, ICSM 2011 29 Sept 2011 Delft University of Technology Challenge the future
- 2. Contributions • Correlation sourcecode metrics vs #changes in interfaces: • C&K metrics • complexity and usage metrics • interface usage cohesion metric • Predictive power of source code metrics for interfaces: • prediction models • 10 open source projects • 8 Eclipse projects • Hibernate 2 and Hibernate 3 2
- 3. Motivations • Changes ininterfaces are not desirable • changes can have stronger impact • interfaces define contracts • existing object oriented metrics not sound for interfaces • Related work about metrics as quality predictors • no differences among the kind of class 3
- 4. Hypotheses • H1 • InterfaceUsageCohesion (IUC) has a stronger correlation with number of Source Code Changes (#SCC) of interfaces than the C&K metrics • H2 • IUC can improve the performance of prediction models to classify Java interfaces into change- and not- change-prone 4
- 5. The Approach source code repository metrics Changes computation Retrieval Spearman rank Correlation Metrics train models correlation Prediction analysis Changes classify interfaces analysis H1 H2 5
- 6. Metrics Computation Evolizer Model source code Importer repository Famix Model Computation Metrics Values Understand 6
- 7. Changes Computation Evolizer source code Version Control repository Connector Revisions Info & Subsequent Changes Computation files Fine-Grained Evolizer Change Source Code Distiller Changes (SCC) AST Comparison 7
- 8. Why SCC? • Filteringout useless changes due to modification of: • licenses • comments • More precise measurement #Revision=1 #LineModified=1 #SCC=2 8
- 9. C&K Correlation forInterfaces Project CB0 NOC RFC DIT LCOM WMC Hibernate3 0.535** 0.029 0.592** 0.058 0.103 0.657** Hibernate2 0.373** 0.065 0.325** -0.01 0.006 0.522** ecl.debug.core 0.484** 0.105 0.486** 0.232* 0.337 0.597** ecl.debug.ui 0.216* 0.033 0.152 0.324** 0.214* 0.131 ecl.jface 0.239* 0.012 0.174** 0.103 0.320** 0.137 ecl.jdt.debug 0.512** 0.256** 0.349** -0.049 0.238** 0.489** ecl.team.core 0.367* 0.102 0.497** 0.243 0.400 0.451** ecl.team.cvs.core 0.688** -0.013 0.738** 0.618** 0.610** 0.744** ecl.team.ui 0.301* -0.003 0.299* -0.103* 0.395** 0.299* update.core 0.499** -0.007 0.381** 0.146 0.482** 0.729** Median 0.428 0.031 0.365 0.124 0.328 0.505 *= significant at α=0.05 **= significant at α=0.01 9
- 10. Weighted Methods perClass (WMC) • ci cyclomatic complexity of the ith method • n number of methods in a class Number of Methods 10
- 11. Interface Segregation Principle ISP defined by Robert C. Martin cope with fat interfaces Fat interface interfaces that serve different clients each kind of client uses a different set of methods the interface should be split in more interface, each one designed to serve a specific client 11
- 12. Interface Segregation Principle(I) Different clients do not share any methods ClusterClients(i): counts the number of clients that do not share any method of the interface i 12
- 13. Interface Usage Cohesion Different clients share a method 13
- 14. Other metrics forinterfaces… • Number Of Methods (NOM) • Number Of Arguments (NOA) • Arguments Per Procedure (APP) • Number of Clients (Cli) • Number of Invocations (Inv) • Number of Implementing Classes (Impl) 14
- 15. Correlation for Interfaces Project Inv Cli NOM Clust IUC Hibernate3 0.544** 0.433** 0.657** 0.302** -0.601** Hibernate2 0.165 0.104 0.522** 0.016 -0.373** ecl.debug.core 0.317** 0.327** 0.597** 0.273** -0.682** ecl.debug.ui 0.497** 0.498** 0.131 0.418** -0.508** ecl.jface 0.205 0.099 0.137 0.106** -0.363** ecl.jdt.debug 0.495** 0.471 0.489** 0.474** -0.605** ecl.team.core 0.261 0.278 0.451** 0.328* -0.475** ecl.team.cvs.core 0.557** 0.608** 0.744** 0.369 -0.819** ecl.team.ui 0.290 0.270 0.299 0.056 -0.618** update.core 0.677** 0.656** 0.729** 0.606** -0.656** Median 0.317 0.327 0.505 0.328 -0.605 *= significant at α=0.05 **= significant at α=0.01 15
- 16. Prediction Analysis • ThreeMachine Learning Algorithms • upport Vector Machine S • aïve Bayes Network N • eural Nets N • Interfaces classification: • Training using 10 fold cross-validation • {CBO, RFC, LCOM, WMC} = CK • {CBO, RFC, LCOM, WMC, IUC} = IUC 16
- 17. Prediction – AUCvalues NBayes LibSVN NN Project CK IUC CK IUC CK IUC ecl.team.cvs.core 0.55 0.75 0.692 0.811 0.8 0.8 ecl.debug.core 0.75 0.79 0.806 0.828 0.85 0.875 ecl.debug.ui 0.66 0.72 0.71 0.742 0.748 0.766 Hibernate2 0.745 0.807 0.735 0.708 0.702 0.747 Hibernate3 0.835 0.862 0.64 0.856 0.874 0.843 ecl.jdt.debug 0.79 0.738 0.741 0.82 0.77 0.762 ecl.jface 0.639 0.734 0.607 0.778 0.553 0.542 ecl.team.core 0.708 0.792 0.617 0.608 0.725 0.85 ecl.team.ui 0.88 0.8 0.74 0.884 0.65 0.75 update.core 0.782 0.811 0.794 0.817 0.675 0.744 Median 0.747 0.791 0.722 0.814 0.736 0.764 17
- 18. Results • H1 ACCEPTED • IUC has a stronger correlation with #SCC of interfaces than the C&K metrics • UIC shows the best correlation • H2 PARTIALLY ACCEPTED • IUC can improve the performance of prediction models to classify Java interfaces into change- and not- change-prone • Despite the improvements Wilcoxon test showed a significant difference only for the LibSVM 18
- 19. Implications • Researchers • taking in account the nature of the measured entities • Quality Engineers • enlarge metrics suites • Developers and Architects • Measure the ISP violation 19
- 20. Future Work • Metrics measurementovertime • Further validation • Are the shared methods the problem? • Component Based System and Service Oriented System 20
- 21.