Detecting Code Smells using Deep Learning

Das, Ananta Kumar; Yadav, Shikhar; Dhal, Subhasish

doi:10.1109/tencon.2019.8929628

Cited by 24 publications

(13 citation statements)

References 7 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…We notice that the accuracy of the 18-layer CNN model is higher than that of the 36-layer CNN model. It should be noted that the accuracy of the 18-layer CNN model is 10% lower than that of the five-layer CNN model reported by Das et al 18 The experimental results show that barely increasing the number of CNN layers does not improve the accuracy and may cause the problem of gradient degradation.…”

Section: Motivationmentioning

confidence: 61%

MARS: Detecting brain class/method code smell based on metric–attention mechanism and residual network

Zhang

Dong

2021

J Software Evolu Process

View full text Add to dashboard Cite

Code smell is the structural design defect that makes programs difficult to understand, maintain, and evolve. Existing works of code smell detection mainly focus on prevalent code smells, such as feature envy, god class, and long method. Few works have been done on detecting brain class/method. Furthermore, existing deep‐learning‐based approaches leverage the CNN model to improve accuracy by barely increasing the number of layers, which may cause a problem of gradient degradation. To this end, this paper proposes a novel approach called MARS to detect brain class/method. MARS improves the gradient degradation by employing an improved residual network. It increases the weight value of those important code metrics to label smelly samples by introducing a metric–attention mechanism. To support the training of MARS, a dataset called BrainCode is generated by extracting more than 270,000 samples from 20 real‐world applications. MARS is evaluated on BrainCode and compared to other machine‐learning‐based and deep‐learning‐based approaches. The experimental results demonstrate that the average accuracy of MARS is 2.01 % higher than that of the existing approaches, which improves state‐of‐the‐art.

show abstract

Section: Motivationmentioning

confidence: 61%

MARS: Detecting brain class/method code smell based on metric–attention mechanism and residual network

Zhang

Dong

2021

J Software Evolu Process

View full text Add to dashboard Cite

show abstract

“…In this respect, we aim at defining the most appropriate tools and data analysis methodologies that may help investigating how static analysis warnings impact the detection of this category of code smells. Last but not least, we plan to systematically assess deep learning methods (Das et al 2019;Liu et al 2019), which might more naturally combine features, given that they act directly on source code.…”

Section: Discussionmentioning

confidence: 99%

On the adequacy of static analysis warnings with respect to code smell prediction

et al. 2022

View full text Add to dashboard Cite

Code smells are poor implementation choices that developers apply while evolving source code and that affect program maintainability. Multiple automated code smell detectors have been proposed: while most of them relied on heuristics applied over software metrics, a recent trend concerns the definition of machine learning techniques. However, machine learning-based code smell detectors still suffer from low accuracy: one of the causes is the lack of adequate features to feed machine learners. In this paper, we face this issue by investigating the role of static analysis warnings generated by three state-of-the-art tools to be used as features of machine learning models for the detection of seven code smell types. We conduct a three-step study in which we (1) verify the relation between static analysis warnings and code smells and the potential predictive power of these warnings; (2) build code smell prediction models exploiting and combining the most relevant features coming from the first analysis; (3) compare and combine the performance of the best code smell prediction model with the one achieved by a state of the art approach. The results reveal the low performance of the models exploiting static analysis warnings alone, while we observe significant improvements when combining the warnings with additional code metrics. Nonetheless, we still find that the best model does not perform better than a random model, hence leaving open the challenges related to the definition of ad-hoc features for code smell prediction.

show abstract

“…A method based on DL was devised [12] to detect two CS, such as Brain Class and Brain Method. To generate complex patterns in the higher layer, one of the DL architectures like the Convolution Neural Networks-1D (CNN) model was used in this method.…”

Section: Literature Surveymentioning

confidence: 99%

Extension of Deep Learning Based Feature Envy Detection for Misplaced Fields and Methods

2022

IJIES

View full text Add to dashboard Cite

Code smells (CS) are a severe symptom in software source code that leads to a serious problem in software maintenance and evolution. Feature Envy (FE) is a type of CS that refers to methods that are misplaced in the source code. Many tools like infusion, JDeodorant and JspIRIT have been used to identify the CS. The conversion of source code measurements into predictions is based on manually constructed heuristics. But, manually selecting good features and constructing optimal heuristics is a challenging issue. So, automatic detection of CS is required. The Deep Learning (DL) methods play an important role in achieving the automatic detection of FE(DLFED). The FE detection approach based on DL is intended to discover misplaced methods. The DL technique automatically identifies the source code characteristics for FE identification as well as automatically improves the difficult mapping structure between the features and their prediction. But, a DL-based automatic detection of CS requires a significant number of features with labeled training datasets to provide better prediction results. Therefore, an extension of the DL-based FE detection technique (EDLFED-FM) is proposed in this article to detect both misplaced methods and fields. The semantic relationship between identifiers in the source code program is used to extract the features of misplaced fields and methods. The new decomposition slice method is proposed to convert the extracted features of misplaced methods and fields into the slicing vector composition. This sliced feature helps deep learners understand the relationship between misplaced methods and fields. This automated method is capable of producing labeled training data for DL techniques-based classifiers without the requirement for human intervention. Finally, the experimental outcomes proved that precision rate of suggested method outperforms the JDeodorant and DLFED methods by 14.49% and 10.29 % for free plane, 14.72 % and 12.2 % for Junit, and 21.84% and 14.68 % for JExcelapi applications. Hence, it has been demonstrated that the predicted misplaced methods and fields are moved to the refactoring class for positive testing of source code program with less training data.

show abstract

Detecting Code Smells using Deep Learning

Cited by 24 publications

References 7 publications

MARS: Detecting brain class/method code smell based on metric–attention mechanism and residual network

MARS: Detecting brain class/method code smell based on metric–attention mechanism and residual network

On the adequacy of static analysis warnings with respect to code smell prediction

Extension of Deep Learning Based Feature Envy Detection for Misplaced Fields and Methods

Contact Info

Product

Resources

About