Bad Smell Detection Using Machine Learning Techniques: A Systematic Literature Review

In the realm of software development, detecting code smells is a critical task for ensuring good code quality. God class code smell specifically has a specific nature associated with a great deal of subjectivity due to the levels of coupling and cohesion associated to it. Automated techniques for code smell detection aim to resolve this subjectivity. Machine learning techniques have shown promising results that tend to improve accuracy and reduce the bias associated with other techniques for God class identification. This is due to their pattern recognition capabilities making them more objective in identifying patterns that indicate code smells. However, current results need to be further improved in terms of both accuracy and generalizability. The challenge in the use of machine learning is not only in selecting the most appropriate technique but also lies in effectively representing source code as input patterns fed to Machine Learning (ML) classifier(s). Code representation plays a pivotal role in encoding source code for ML algorithms. This study aims improving the accuracy and generalizability of God class code smell detection via exploring the effect of using various code representation techniques, namely, tree-based, metric-based, code embedding, and token-based code representation techniques on the ML detection results. The study is conducted on the MLCQ dataset, and applies various ML algorithms (specifically: Logistic Regression, Random Forest, SVM, Decision Tree, Naive Bayes, Gradient Boosting, XGBoost). The evaluation results show how different code representation techniques influence ML detection outcomes and the comparative performance of ML algorithms. The study findings reveal that the F1-score achieved outperforms prior studies on the MLCQ dataset, indicating the effectiveness of the proposed approach. The presented results reveal how the code representation technique used makes a significant impact on the ML classifier results. This paves the way for developing intelligent IDE plugins for just in time God Class code smell detection among other code smells.

Background Code smells are an indicator that something is misplaced in software systems, which can reduce maintainability and quality of software applications. In the literature, there is much research about code smells where numerous strategies for automating code smell detection have been developed to improve software quality. Objective The purpose of this work is to provide a review of search‐based, heuristic‐based, machine learning‐based, deep learning‐based, and hybrid‐based code smell detection algorithms. Method Concerning the main goals of this research, we have found 38 primary studies. We gathered relevant studies published on this topic between 2017 and 2024. These articles' data were extracted according to some criteria, including code smells, machine learning methods, programming languages, dataset size, evaluation strategy, and statistical tests. Results Machine learning‐based code smell detection methods have been suggested in most empirical investigations. We found that machine learning and deep learning are the most popular approaches for predicting code smells. Researchers typically employ support vector machine and decision tree algorithms. The Arcelli Fontana and Zanoni benchmark dataset was the most frequently investigated dataset. Most of the research community's attention has been focused on a small number of smells, including blob, feature envy, long method, and data class. Researchers also pay more attention to code smells like Long Method and Feature Envy. Deep learning techniques are increasingly used, and most scientists utilize source code metrics as indicators. The standard performance measurements mostly used are F‐measure, recall, and precision. Conclusions The study provides an overview of existing approaches and highlights current research trends in code smell detection, particularly the increasing use of machine learning and deep learning techniques.

Code smells are considered symptoms of poor implementation choices, which may hamper the software maintainability. Hence, code smells should be detected as early as possible to avoid software quality degradation. Unfortunately, detecting code smells is not a trivial task. Some preliminary studies investigated and concluded that machine learning (ML) techniques are a promising way to better support smell detection. However, these techniques are hard to be customized to promote an early and accurate detection of specific smell types. Yet, ML techniques usually require numerous code examples to be trained (composing a relevant dataset) in order to achieve satisfactory accuracy. Unfortunately, such a dependency on a large validated dataset is impractical and leads to late detection of code smells. Thus, a prevailing challenge is the early customized detection of code smells taking into account the typical limited training data. In this direction, this paper reports a study in which we collected code smells, from ten active projects, that were actually refactored by developers, differently from studies that rely on code smells inferred by researchers. These smells were used for evaluating the accuracy regarding early detection of code smells by using seven ML techniques. Once we take into account such smells that were considered as important by developers, the ML techniques are able to customize the detection in order to focus on smells observed as relevant in the investigated systems. The results showed that all the analyzed techniques are sensitive to the type of smell and obtained good results for the majority of them, especially JRip and Random Forest. We also observe that the ML techniques did not need a high number of examples to reach their best accuracy results. This finding implies that ML techniques can be successfully used for early detection of smells without depending on the curation of a large dataset.

Code smells are specific patterns or characteristics in software code that indicate potential design or implementation problems. Identifying code smells has gained significant attention in software engineering. It is essential to address code smells to maintain high-quality software systems. Machine learning (ML) models, such as Long Short-Term Memory (LSTM), have been to detect code smells automatically based on source code features. However, the imbalanced distribution of code smells within software projects poses a challenge to the accuracy of these models. This study explores the role of data balancing methods in optimizing the accuracy of the LSTM model for code smell detection. We investigate different techniques for addressing the class imbalance problem, including random oversampling and synthetic minority oversampling techniques (SMOTE). We evaluate the performance of the LSTM model with and without data balancing methods using accuracy, precision, recall, f-measure, Matthew’s correlation coefficient (MCC), and the area under a receiver operating characteristic curve (AUC). Our experimental results, conducted on four code smell datasets (God class, data class, feature envy, and long method) extracted from 74 open-source systems, demonstrate the effectiveness of data balancing methods in improving the accuracy of the LSTM model for code smell detection. The results indicate that the use of data balancing methods had a positive effect on the predictive accuracy of the LSTM model. In addition, we compared our proposed method with state-of-the-art code smell detection approaches. The findings from the comparison indicate that our proposed method performs notably better than existing state-of the-art approaches across the majority of datasets.

Code smells are generally not considered as bugs; instead, they point out certain shortcomings in the software design or code. Identification of code smell is a necessary step for improving the software quality and reducing the maintenance effort. In this study, we introduce a bad smell prediction technique based on object-oriented software metrics that use Decision Tree (DT) and Random Forest (RF) machine learning algorithm. An open-source project, namely JHOTDRAW, was used as our dataset, for which values of object-oriented software metrics were calculated. Two feature selection methods-Random Forest Importance (RFI) and Information Gain (IG) were applied to extract the most relevant attributes for the prediction of code smells, namely, Feature envy, Dispersed coupling, refused parent bequest, and God class. The random-search algorithm was used to tune the parameters of Random Forest and Decision Tree. Results show that the best classification accuracy for Decision Tree was obtained at 99.13% for refused parent bequest code smells. Results also show that after using the Random Forest classifier, refused parent bequest smell was predicted with the best accuracy of 99.14%. Finally, in this research study, a set of code smell prediction rules were extracted using Decision Tree.

The concept of code smells is widespread in Software Engineering. Despite the empirical studies addressing the topic, the set of context-dependent issues that impacts the human perception of what is a code smell has not been studied in depth. We call this the code smell conceptualization problem. To discuss the problem, empirical studies are necessary. In this work, we focused on conceptualization of god class. God class is a code smell characterized by classes that tend to centralize the intelligence of the system. It is one of the most studied smells in software engineering literature. A controlled experiment that extends and builds upon a previous empirical study about how humans detect god classes, their decision drivers, and agreement rate. Our study delves into research questions of the previous study, adding visualization to the smell detection process, and analyzing strategies of detection. Our findings show that agreement among participants is low, which corroborates previous studies. We show that this is mainly related to agreeing on what a god class is and which thresholds should be adopted, and not related to comprehension of the programs. The use of visualization did not improve the agreement among the participants. However, it did affect the choice of detection drivers. This study contributes to expand empirical evidences on the impact of human perception on detecting code smells. It shows that studies about the human role in smell detection are relevant and they should consider the conceptualization problem of code smells.

Code smell detection is essential to improve software quality, enhancing software maintainability, and decrease the risk of faults and failures in the software system. In this paper, we proposed a code smell prediction approach based on machine learning techniques and software metrics. The local interpretable model-agnostic explanations (LIME) algorithm was further used to explain the machine learning model’s predictions and interpretability. The datasets obtained from Fontana et al. were reformed and used to build binary-label and multi-label datasets. The results of 10-fold cross-validation show that the performance of tree-based algorithms (mainly Random Forest) is higher compared with kernel-based and network-based algorithms. The genetic algorithm based feature selection methods enhance the accuracy of these machine learning algorithms by selecting the most relevant features in each dataset. Moreover, the parameter optimization techniques based on the grid search algorithm significantly enhance the accuracy of all these algorithms. Finally, machine learning techniques have high potential in predicting the code smells, which contribute to detect these smells and enhance the software’s quality.

The field of software refactoring has been evolving rapidly, driven by the need to maintain software quality and manage complexity as software systems grow. This special issue of the Journal of Software: Evolution and Process focuses on “Software Refactoring: Application Breadth and Technical Depth,” aiming to promote research that advances our understanding of software refactoring practices and their impact. The selected papers in this issue reflect the broad scope and technical depth of the topic, covering a variety of innovative methodologies, empirical studies, and new tools that push the boundaries of current software refactoring research. The paper “Software Refactoring Side Effects” by AbuHassan, Alshayeb, and Ghouti addresses the often-overlooked side effects of software refactoring. The authors propose three different approaches for handling these side effects and employ a multi-objective optimization algorithm to identify optimal refactoring decisions. Their work enhances the understanding of refactoring's unintended consequences and proposes practical solutions, fitting well into the theme of exploring new dimensions of refactoring practices. Zhao et al., in their article “An Architecture Refactoring Approach to Reducing Software Hierarchy Complexity,” introduce a novel metric to quantify software hierarchy complexity and propose the Complexity-oriented Software Architecture Refactoring (CoSSR) approach. Their study demonstrates how reducing software hierarchy complexity can improve software quality, aligning with the issue's focus on reducing complexity through innovative refactoring techniques. The article “MARS: Detecting Brain Class/Method Code Smell Based on Metric-Attention Mechanism and Residual Network” by Zhang and Dong presents a new method for detecting complex code smells using a novel deep learning approach. By leveraging a metric-attention mechanism and an improved residual network, their model shows superior performance compared to existing methods. This research contributes to the special issue by applying cutting-edge machine learning techniques to refactoring challenges. AlOmar et al.'s study, “Behind the Scenes: On the Relationship Between Developer Experience and Refactoring,” investigates the correlation between developer experience and refactoring activities across 800 open-source projects. Their findings highlight that experienced developers tend to perform more varied refactoring operations but document less of their refactoring activities. This study offers a new perspective on how human factors influence refactoring practices, enriching the broader conversation on effective software evolution. In “A Study of Refactorings During Software Change Tasks,” Eilertsen and Murphy explore the reasons behind the disuse of refactoring tools during software change tasks. Their observational study reveals that the strategies developers use significantly impact the use of refactoring tools, pointing to a gap between available tool support and actual developer needs. This paper provides valuable insights into the practical challenges of implementing refactoring in real-world settings. Halepmollasi and Tosun, in “Exploring the Relationship Between Refactoring and Code Debt Indicators,” present an empirical analysis of the association between different types of refactorings, code smells, and faults. Their findings challenge previous assumptions about refactoring's impact on code smells and suggest that refactoring activities might be more fault-inducing than previously thought. This research adds a critical dimension to understanding the effects of refactoring on software quality. Lastly, Abdou and Darwish's article, “Severity Classification of Software Code Smells using Machine Learning Techniques: A Comparative Study,” evaluates various machine learning models for classifying the severity of code smells. Their use of the LIME algorithm for model interpretability provides new insights into predicting code smell severity, contributing to the issue's exploration of data-driven approaches to software maintenance. Recent developments in generative AI and machine learning have opened new avenues for advancing software refactoring practices. As demonstrated by some of the studies in this issue, these technologies can automate and enhance refactoring processes, improve tool support, and provide more accurate code smell detection. The application of AI in this domain is still in its early stages, but it holds great promise for transforming how software systems are maintained and evolved. We encourage researchers to explore these emerging trends further, focusing on integrating AI-driven approaches with traditional software engineering practices. We extend our sincere appreciation to all the authors for their contributions, to the peer reviewers for their invaluable feedback and dedication, and to the editorial board members for their support in assembling this special issue. Together, these efforts have created a diverse collection of research that we believe will significantly impact the field of software refactoring.

Code smells can be subjectively interpreted, the results provided by detectors are usually different, the agreement in the results is scarce, and a benchmark for the comparison of these results is not yet available. The main approaches used to detect code smells are based on the computation of a set of metrics. However code smell detectors often use different metrics and/or different thresholds, according to their detection rules. As result of this inconsistency the number of detected smells can increase or decrease accordingly, and this makes hard to understand when, for a specific software, a certain characteristic identifies a code smell or not. In this work, we introduce WekaNose, a tool that allows to perform an experiment to study code smell detection through machine learning techniques. The experiment's purpose is to select rules, and/or obtain trained algorithms, that can classify an instance (method or class) as affected or not by a code smell. These rules have the main advantage of being extracted through an example-based approach, rather then a heuristic-based one.

The quality of source code is negatively impacted by code smells. Since the term "code smell" originated, numerous attempts have been made to comprehend it by identifying it using various techniques, such as metric-based, heuristic-based, optimization-based, machine learning (ML)-based, etc. Among these, supervised machine learning (SML) has shown effectiveness in detecting code smells. However, SML techniques have significant limitations, including the dependency on expensive and high-quality labeled data, the need for representative training datasets, and the risk of introducing biases in labeled examples that lead to skewed predictions. To overcome these challenges, this study introduces a method that leverages unsupervised machine learning (UnML) along with feature engineering. Unlike SML, UnML does not require labeled data and minimizes potential biases. The proposed method was evaluated using four datasets containing different types of code smells and was compared with a previous study that used SML techniques. The results indicate that the UnML-based method is effective, achieving outcomes closely aligned with those from the SML approach. This method is especially beneficial in situations where labeled data is scarce or unavailable and can be used to identify new code smells, generate labeled data for SML and detect multiple code smells simultaneously within a codebase.

Improving and comparing performance of machine learning classifiers optimized by swarm intelligent algorithms for code smell detection

ObjectiveEnhanced technology in computer and internet has driven scale and quality of data to be improved in various areas including healthcare sectors. Machine Learning (ML) has played a pivotal role in efficiently analyzing those big data, but a general misunderstanding of ML algorithms still exists in applying them (e.g., ML techniques can settle a problem of small sample size, or deep learning is the ML algorithm). This paper reviewed the research of diagnosing mental illness using ML algorithm and suggests how ML techniques can be employed and worked in practice.MethodsResearches about mental illness diagnostic using ML techniques were carefully reviewed. Five traditional ML algorithms-Support Vector Machines (SVM), Gradient Boosting Machine (GBM), Random Forest, Naïve Bayes, and K-Nearest Neighborhood (KNN)-frequently used for mental health area researches were systematically organized and summarized.ResultsBased on literature review, it turned out that Support Vector Machines (SVM), Gradient Boosting Machine (GBM), Random Forest, Naïve Bayes, and K-Nearest Neighborhood (KNN) were frequently employed in mental health area, but many researchers did not clarify the reason for using their ML algorithm though every ML algorithm has its own advantages. In addition, there were several studies to apply ML algorithms without fully understanding the data characteristics.ConclusionResearchers using ML algorithms should be aware of the properties of their ML algorithms and the limitation of the results they obtained under restricted data conditions. This paper provides useful information of the properties and limitation of each ML algorithm in the practice of mental health.

Are you smelling it? Investigating how similar developers detect code smells

Code Smells have been detected, predicted and studied by researchers from several perspectives. This literature review is conducted to understand tools and algorithms used to detect and analyze code smells to summarize research agenda. 114 studies have been selected from 2009 to 2022 to conduct this review. The studies are deeply analyzed under the categorization of machine learning and non-machine learning, which are found to be 25 and 89 respectively. The studies are analyzed to gain insight into algorithms, tools and limitations of the techniques. Long Method, Feature Envy, and Duplicate Code are reported to be the most popular smells. 38% of the studies focused their research on the enhancement of tools and methods. Random Forest and JRip algorithms are found to give the best results under machine learning techniques. We extended the previous studies on code smell detection tools, reporting a total 87 tools during the review. Java is found to be the dominant programming language during the study of smells.

Code smells are code structures that indicate a potential issue in code design or implementation. These issues could affect the processes of code testing and maintenance, and overall software quality. Therefore, it is important to detect code smells in the early stages of software development to enhance system quality. Most studies have focused on detecting code smells of a single programming language. This article explores TL for cross-language code smell detection, where Java is the source, and both C# and Python are the target datasets, focusing on Large Class, Long Method, and Long Parameter List code smells. We conducted a comparison study across two transfer learning approaches—instance-based (Importance Weighting Classifier, Nearest Neighbors Weighting, and Transfer AdaBoost) and parameter-based (Transfer Tree, Transfer Forest)—with various base models. The results showed that the instance-based approach outperformed the parameter-based approach, particularly with Transfer AdaBoost using ensemble learning base models. The Transfer AdaBoost approach with Gradient Boosting and Extra Trees achieved consistent and robust results across both C# and Python, with an 83% winning rate, as indicated by the Wilcoxon signed-rank test. These findings underscore the effectiveness of transfer learning for cross-language code smell detection, supporting its generalizability across different programming languages.