Comparing Heuristic and Machine Learning Approaches for Metric-Based Code Smell Detection

Pecorelli, Fabiano; Palomba, Fabio; Nucci, Dario Di; Lucia, Andrea De

doi:10.1109/icpc.2019.00023

Cited by 83 publications

(51 citation statements)

References 82 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…According to several empirical studies [88,89,100], the accuracy of Decor is relatively high both in terms of precision and recall, with typical values of F-Measure around 75%. This makes the detector more accurate than other available tools [41] and, therefore, suitable for our study.…”

Section: Styleandidementioning

confidence: 99%

On the performance of method-level bug prediction: A negative result

Pascarella

Palomba

Bacchelli

2020

Journal of Systems and Software

Self Cite

View full text Add to dashboard Cite

Bug prediction is aimed at identifying software artifacts that are more likely to be defective in the future. Most approaches defined so far target the prediction of bugs at class/file level. Nevertheless, past research has provided evidence that this granularity is too coarse-grained for its use in practice. As a consequence, researchers have started proposing defect prediction models targeting a finer granularity (particularly method-level granularity), providing promising evidence that it is possible to operate at this level. Particularly, models mixing product and process metrics provided the best results. We present a study in which we first replicate previous research on method-level bug-prediction, by using different systems and timespans. Afterwards, based on the limitations of existing research, we (1) re-evaluate method-level bug prediction models more realistically and (2) analyze whether alternative features based on textual aspects, code smells, and developer-related factors can be exploited to improve method-level bug prediction abilities. Key results of our study include that (1) the performance of the previously proposed models, tested using the same strategy but on different systems/timespans, is confirmed; but, (2) when evaluated with a more practical strategy, all the models show a dramatic drop in performance, with results close to that of a random classifier. Finally, we find that (3) the contribution of alternative features within such models is limited and unable to improve the prediction capabilities significantly. As a consequence, our replication and negative results indicate that method-level bug prediction is still an open challenge.

show abstract

Section: Styleandidementioning

confidence: 99%

On the performance of method-level bug prediction: A negative result

Pascarella

Palomba

Bacchelli

2020

Journal of Systems and Software

Self Cite

View full text Add to dashboard Cite

show abstract

“…Method PV Generic a JHD b TP TriangleFigure.polygon() 74 ---------ShortestDistanceConnector.findPoint(..) 34 14 --------AlignCommand.execute() 31…”

Section: I4) How High Is Our Confidence In Our Bad Smell Improvement?mentioning

confidence: 99%

“…(D6) "Known Type in Aggregation" The last line of drawHandles(Graphics) in the class StandardDrawingView presented in table 4 takes an object from the enumeration and invokes the method draw(Graphics) on it, which requires casting the object to Handle. 31 StandardDrawingView has a one to many association to the class Handle, which is realized through the field fSelectionHandles of type Vector.…”

Section: :41mentioning

confidence: 99%

See 1 more Smart Citation

Did JHotDraw respect the Law of Good Style? - An exploratory deep dive into the nature of false positives of bad code smells

Speicher

2020

Programming

View full text Add to dashboard Cite

Developers need to make a constant effort to improve the quality of their code if they want to stay productive. Tools that highlight code locations that could benefit from refactoring are thus highly desirable. The most common name for such locations is "bad code smell". A number of tools offer such quality feedback and there is a substantial body of related research.However, all these tools, including those based on Machine Learning, still produce false positives. Every single false positive shown to the developer places a cognitive burden on her and should thus be avoided. The literature discusses the choice of metric thresholds, the general subjectivity of such a judgment and the relation to conscious design choices, "design ideas".To examine false positives and the relation between bad smells and design ideas, we designed and conducted an exploratory case study. While previous research presented a broad overview, we have chosen a narrow setting to reach for even deeper insights: The framework JHotDraw had been designed so thoughtfully that most smell warnings are expected to be false positives. Nevertheless, the "Law of Good Style", better known as the "Law of Demeter", is a rather restrictive design rule so that we still expected to find some potential bad smells, i.e. violations of this "Law".This combination led to 1215 potential smells of which at most 42 are true positives. We found generic as well as specific design ideas that were traded for the smell. Our confidence in that decision ranged from high enough to very high. We were surprised to realize that the smell definition itself required the formulation of constructive design ideas. Finally we found some smells to be the result of the limitation of the language and one could introduce auxiliary constructive design ideas to compensate for them.The decision whether a potential smell occurrence is actually a true positive was made very meticulously. For that purpose we took three qualities that the smell could affect negatively into account and we discussed the result of the recommended refactorings. If we were convinced that we had found a false positive, we described the relationships with design ideas.The realization that not only general design ideas but also specific design ideas have an influence on whether a potential smell is a true positive turns the problem of false positives from a scientific problem ("What is the true definition of the smell?") to a engineering problem ("How can we incorporate design ideas into smell definitions?"). We recommend to add adaptation points to the smell definitions. Higher layers may then adapt the smell for specific contexts. After adaptation the tool may continuously provide distinct and precise quality feedback, reducing the cognitive load for the developer and preventing habituation. Furthermore, the schema for the discussion of potential smells may be used to elaborate more sets of true and false smell occurrences. Finally, it follows that smell detection based on machine learning should also take signs of...

show abstract

“…For supporting smell identification, several detection tools have been proposed [4] [12]. However, even though these tools may save developers' effort on identification, they cannot be considered the final word [10].…”

Section: Introductionmentioning

confidence: 99%

Revealing Developers’ Arguments on Validating the Incidence of Code Smells: A Focus Group Experience

Junionello¹,

Mello²,

Oliveira³

et al. 2021

Anais Do IX Workshop De Visualização, Evolução E Manutenção De Software (VEM 2021)

View full text Add to dashboard Cite

Identifying code smells is considered a subjective task. Unfortunately, current automated detection tools cannot deal with such subjectivity, requiring human validation. Developers tend to follow different, albeit complementary, strategies when validating the identified smells. Intending to find out developers' arguments when validating the incidence of code smells, we conducted a focus group session with developers familiar with identifying code smells. We distributed them among two groups, in which they had to argue about the incidence of a code smell: either accepting or rejecting its presence. Based on their arguments, we compiled a set of general heuristics that developers follow when validating smells. We then used these heuristics for composing validation items. We understand that the set of validation items proposed may support developers in reflecting on the incidence of code smells. However, further studies are needed for reaching a more comprehensive and optimized set. The experience of this study reveals that conducting focus group sessions is helpful to emerge the tacit knowledge of developers when validating code smells.

show abstract

Comparing Heuristic and Machine Learning Approaches for Metric-Based Code Smell Detection

Cited by 83 publications

References 82 publications

On the performance of method-level bug prediction: A negative result

On the performance of method-level bug prediction: A negative result

Did JHotDraw respect the Law of Good Style? - An exploratory deep dive into the nature of false positives of bad code smells

Revealing Developers’ Arguments on Validating the Incidence of Code Smells: A Focus Group Experience

Contact Info

Product

Resources

About