Software Fault Research Articles

Improving accuracy of code smells detection using machine learning with data balancing techniques

Code smells indicate potential symptoms or problems in software due to inefficient design or incomplete implementation. These problems can affect software quality in the long-term. Code smell detection is fundamental to improving software quality and maintainability, reducing software failure risk, and helping to refactor the code. Previous works have applied several prediction methods for code smell detection. However, many of them show that machine learning (ML) and deep learning (DL) techniques are not always suitable for code smell detection due to the problem of imbalanced data. So, data imbalance is the main challenge for ML and DL techniques in detecting code smells. To overcome these challenges, this study aims to present a method for detecting code smell based on DL algorithms (Bidirectional Long Short-Term Memory (Bi-LSTM) and Gated Recurrent Unit (GRU)) combined with data balancing techniques (random oversampling and Tomek links) to mitigate data imbalance issue. To establish the effectiveness of the proposed models, the experiments were conducted on four code smells datasets (God class, data Class, feature envy, and long method) extracted from 74 open-source systems. We compare and evaluate the performance of the models according to seven different performance measures accuracy, precision, recall, f-measure, Matthew’s correlation coefficient (MCC), the area under a receiver operating characteristic curve (AUC), the area under the precision–recall curve (AUCPR) and mean square error (MSE). After comparing the results obtained by the proposed models on the original and balanced data sets, we found out that the best accuracy of 98% was obtained for the Long method by using both models (Bi-LSTM and GRU) on the original datasets, the best accuracy of 100% was obtained for the long method by using both models (Bi-LSTM and GRU) on the balanced datasets (using random oversampling), and the best accuracy 99% was obtained for the long method by using Bi-LSTM model and 99% was obtained for the data class and Feature envy by using GRU model on the balanced datasets (using Tomek links). The results indicate that the use of data balancing techniques had a positive effect on the predictive accuracy of the models presented. The results show that the proposed models can detect the code smells more accurately and effectively.

ULTRA-Effective Labeling of Repetitive Genomic Sequence.

In the age of long read sequencing, genomics researchers now have access to accurate repetitive DNA sequence (including satellites) that, due to the limitations of short read sequencing, could previously be observed only as unmappable fragments. Tools that annotate repetitive sequence are now more important than ever, so that we can better understand newly uncovered repetitive sequences, and also so that we can mitigate errors in bioinformatic software caused by those repetitive sequences. To that end, we introduce the 1.0 release of our tool for identifying and annotating locally-repetitive sequence, ULTRA (ULTRA Locates Tandemly Repetitive Areas). ULTRA is fast enough to use as part of an efficient annotation pipeline, produces state-of-the-art reliable coverage of repetitive regions containing many mutations, and provides interpretable statistics and labels for repetitive regions. It released under an open license, and available for download at https://github.com/TravisWheelerLab/ULTRA.

Making More with Less: Improving Software Testing Outcomes Using a Cross-Project and Cross-Language ML Classifier Based on Cost-Sensitive Training

As digitalization expands across all sectors, the economic toll of software defects on the U.S. economy reaches up to $2.41 trillion annually. High-profile incidents like the Boeing 787-Max 8 crash have shown the devastating potential of these defects, highlighting the critical importance of software testing within quality assurance frameworks. However, due to its complexity and resource intensity, the exhaustive nature of comprehensive testing often surpasses budget constraints. This research utilizes a machine learning (ML) model to enhance software testing decisions by pinpointing areas most susceptible to defects and optimizing scarce resource allocation. Previous studies have shown promising results using cost-sensitive training to refine ML models, improving predictive accuracy by reducing false negatives through addressing class imbalances in defect prediction datasets. This approach facilitates more targeted and effective testing efforts. Nevertheless, these models’ in-company generalizability across different projects (cross-project) and programming languages (cross-language) remained untested. This study validates the approach’s applicability across diverse development environments by integrating various datasets from distinct projects into a unified dataset, using a more interpretable ML technique. The results demonstrate that ML can support software testing decisions, enabling teams to identify up to 7× more defective modules compared to benchmark with the same testing effort.

Deep learning or classical machine learning? An empirical study on line‐level software defect prediction

Deep learning or classical machine learning? An empirical study on line‐level software defect prediction

Research on Machine Learning Program Generation Algorithm Based on AORBCO

Abstract The design and development of machine learning programs require selecting appropriate data and algorithms, and coding and debugging based on specific task requirements and the programming experience of developers. However, the current knowledge structure in the field of machine learning is relatively complex, lacking systematic organization, and developers often face the problem of lack of experience when choosing algorithms and designing programs, resulting in a long development cycle and easy errors in machine learning programs. In response to the above issues, this article proposes and designs a machine learning program generation algorithm based on the AORBCO model. The program generation ability includes two sub abilities: algorithm decision-making ability and code generation ability. AD-EKG has been designed for algorithmic decision-making ability, allowing Ego to select appropriate machine learning algorithms based on datasets in massive data. This algorithm combines the characteristics of the AORBCO model's domain knowledge base, knowledge graph based recommendation algorithm, and collaborative filtering algorithm. By calculating the descriptive and structural information between the dataset and algorithm, the interaction probability between the dataset and algorithm is obtained, allowing Ego to make algorithmic decisions interaction probability based. Results of the experiment have shown that the AD-EKG algorithm can fully utilize structural and descriptive information to improve the accuracy of Ego algorithm decision-making. CodeT5-EKG has been designed for code generation capability, allowing Ego to automatically generate machine learning program code. This algorithm combines the CodeT5 generative model with the domain knowledge base of the AORBCO model, by adding auxiliary information extracted using DPR technology to the code generation task, and performing diversified fusion operations to improve code generation quality. The CodeT5-EKG algorithm combines the creativity and efficiency of generative models and DPR technology, and is an algorithm that can improve the quality of generated code while also having the advantages of generative models. The experiments have proved that the code generated by this algorithm has better quality compared to other generative models with the same number of parameters.

Characterizing the safety profile of vagus nerve stimulation devices for epilepsy from 21,448 manufacturer and user reports.

This study summarizes medical device reports (MDRs) associated with adverse events for vagus nerve stimulation (VNS) devices indicated for epilepsy as reported by the Manufacturer and User Facility Device Experience (MAUDE) database of the US Food and Drug Administration. The MAUDE database was surveyed for MDRs from November 2013 to September 2022 regarding VNS devices for epilepsy. Event descriptions, device problems, correlated patient consequences, and device models were grouped and analyzed in Python. Based on event description, revision surgeries and other unique events were identified. Revenue from VNS device sales was used to approximate growth in their use over time. A total of 21,448 MDRs met the inclusion criteria. High VNS impedance, the most prevalent device malfunction overall (17.0% of MDRs), was the most common factor for 18 of the 102 encountered patient problems and led to 1001 revision surgeries (3371 total revisions). Included in those 18 device malfunctions were 3 of the top 6 occurring patient problems: seizure recurrence (9.9% associated with high impedance; encompassed focal, absence, and grand mal subtypes), death (1.3%), and generalized pain (7.9%). The next 4 top cited device malfunctions-lead fracture (13.7% of MDRs), operational issue (6.6%), battery problem holding charge (4.2%), and premature end-of-life indicator (2.9%)-differed widely in their percentage of cases that did not impact patients (77.4%, 57.3%, 48.9%, and 92.2%, respectively), highlighting differing malfunction severities. Seizure recurrence, the most prevalent patient impact, was the outcome most associated with 32 of the 68 encountered device problems, including high impedance (12.8%), lead fracture (12.2%), operational issue (18.4%), battery problem holding charge (31.2%), and premature end-of-life indicator (8.9%), which comprised the top 5 occurring device problems. In general, MDRs spanned a diverse range including device age, hardware, software, and surgeon or manufacturer error. Trends were seen over time with declining annual MDRs coupled with a rise in the use of VNS devices as gauged by revenue growth. Shifting device and patient problem profiles were also seen in successive models, reflecting engineering updates. This study characterizes the most common and consequential side effects of VNS devices for epilepsy while clarifying likely causes. In addition, the outcomes of 68 distinct device malfunctions were identified, including many not ubiquitously present in literature, lending critical perspective to clinical practice.

Optimized multi correlation-based feature selection in software defect prediction

Optimized multi correlation-based feature selection in software defect prediction

Cross-project software defect prediction through multiple learning

Cross-project defect prediction is a method that predicts defects in one software project by using the historical record of another software project. Due to distribution differences and the weak classifier used to build the prediction model, this method has poor prediction performance. Cross-project defect prediction may perform better if distribution differences are reduced, and an appropriate individual classifier is chosen. However, the prediction performance of individual classifiers may be affected in some way by their weaknesses. As a result, in order to boost the accuracy of cross-project defect prediction predictions, this study proposed a strategy that makes use of multiple classifiers and selects attributes that are similar to one another. The proposed method's efficacy was tested using the Relink and AEEEM datasets in an experiment. The findings of the experiments demonstrated that the proposed method produces superior outcomes. To further validate the method, we employed the Wilcoxon sum rank test at 95% significance level. The approach was found to perform significantly better than the baseline methods.

Software Defect Prediction Based on Deep Representation Learning of Source Code From Contextual Syntax and Semantic Graph

Software Defect Prediction Based on Deep Representation Learning of Source Code From Contextual Syntax and Semantic Graph

A software defect prediction method using binary gray wolf optimizer and machine learning algorithms

ContextSoftware defect prediction means finding defect-prone modules before the testing process which will reduce testing cost and time. Machine learning methods can provide valuable models for developers to classify software faulty modules. ProblemThe inherent problem of the classification is the large volume of the training dataset's features, which reduces the accuracy and precision of the classification results. The selection of the effective features of the training dataset for classification is an NP-hard problem that can be solved using heuristic algorithms. MethodIn this study, a binary version of the Gray Wolf optimizer (bGWO) was developed to select the most effective features of the training dataset. By selecting the most influential features in the classification, the precision and accuracy of the software module classifiers can be increased. ContributionDeveloping a binary version of the gray wolf optimization algorithm to optimally select the effective features and creating an effective defect predictor are the main contributions of this study. To evaluate the effectiveness of the proposed method, five real-world and standard datasets have been used for the training and testing stages of the classifier. ResultsThe results indicate that among the 21 features of the train datasets, the basic complexity, sum of operators and operands, lines of codes, number of lines containing code and comments, and sum of operands have the greatest effect in predicting software defects. In this research, by combining the bGWO method and machine learning algorithms, accuracy, precision, recall, and F1 criteria have been considerably increased.

Data-efficient software defect prediction: A comparative analysis of active learning-enhanced models and voting ensembles

As software systems undergo escalating complexity, the identification of bugs and defects becomes pivotal for ensuring seamless user experiences and averting potentially costly post-release issues. This study addresses this critical need by concentrating on the application of active learning methods in code defect prediction. The investigation focuses on the efficacy of active learning combined with ensemble methods, leveraging the dynamic selection and labeling of training instances to increase model performance, reduce the demand for exhaustive labeling efforts, and enhance the effectiveness of code defect prediction systems. Various traditional and ensemble methods are deployed, employing diverse query strategies (uncertainty, margin, and entropy sampling) to assess if active variants can rival original approaches while significantly downsizing the training set. Evaluation encompasses classical classification metrics (AUC, Kappa, and MCC), supplemented by a proposed easy-to-interpret performance index that takes into consideration not only the traditional metric outcomes but also the percentage of the initial dataset utilized, aligned with the dual nature of the problem. The results, elucidated through graphical representations and statistical tests, unveil the advantage of active methods, showcasing reductions of the initial training set by at least 75% in approximately 64% of cases.

Hybrid feature selection method for predicting software defect

To address the challenges associated with the abundance of features in software datasets, this study proposes a novel hybrid feature selection method that combines quantum particle swarm optimization (QPSO) and principal component analysis (PCA). The objective is to identify a subset of relevant features that can effectively contribute to the accuracy of a predictive model based on an artificial neural network (ANN). The quantum particle swarm optimization algorithm is employed to optimize the selection of features by simulating the behavior of quantum particles in a search space. This approach enhances the exploration and exploitation capabilities, allowing for a more effective identification of relevant features. Furthermore, principal component analysis is integrated into the hybrid method to reduce dimensionality and remove multicollinearity among features, thereby improving the efficiency of the feature selection process. The proposed hybrid method is applied to software defect datasets, where the selected subset of features is fed into an artificial neural network for defect prediction. The performance of the hybrid model is compared with traditional feature selection methods, standalone QPSO, and PCA. Experimental results demonstrate the effectiveness of the hybrid approach in achieving superior predictive accuracy while reducing the dimensionality of the dataset. The proposed approach not only enhances prediction accuracy but also provides a more interpretable and efficient subset of features for building robust defect prediction models.

Software Defect Prediction Using Deep Q-Learning Network-Based Feature Extraction

Machine learning-based software defect prediction (SDP) approaches have been commonly proposed to help to deliver high-quality software. Unfortunately, all the previous research conducted without effective feature reduction suffers from high-dimensional data, leading to unsatisfactory prediction performance measures. Moreover, without proper feature reduction, the interpretability and generalization ability of machine learning models in SDP may be compromised, hindering their practical utility in diverse software development environments. In this paper, an SDP approach using deep Q-learning network (DQN)-based feature extraction is proposed to eliminate irrelevant, redundant, and noisy features and improve the classification performance. In the data preprocessing phase, the undersampling method of BalanceCascade is applied to divide the original datasets. As the first step of feature extraction, the weight ranking of all the metric elements is calculated according to the expected cross-entropy. Then, the relation matrix is constructed by applying random matrix theory. After that, the reward principle is defined for computing the Q value of Q-learning based on weight ranking, relation matrix, and the number of errors, according to which a convolutional neural network model is trained on datasets until the sequences of metric pairs are generated for all datasets acting as the revised feature set. Various experiments have been conducted on 11 NASA and 11 PROMISE repository datasets. Sensitive analysis experiments show that binary classification algorithms based on SDP approaches using the DQN-based feature extraction outperform those without using it. We also conducted experiments to compare our approach with four state-of-the-art approaches on common datasets, which show that our approach is superior to these methods in precision, F-measure, area under receiver operating characteristics curve, and Matthews correlation coefficient values.

Analysis of Field Inspection Issues and Some Suggestions for Software as Medical Device

With the encouragement of policies and the rapid development of the biopharmaceutical industry, the number of software as medical device (SaMD) registration applications in Shanghai has continued to increase in recent years, and this paper summarizes the GMP nonconformities found in the field inspection of SaMD in Shanghai from 2020 to 2023, and the results show that nearly 70% of the problems were found in the software development process. Through in-depth analysis, this paper proposes the corresponding countermeasures for the problems found in the five most common stages such as software requirements, software design, software testing, software defect management and software configuration management, combined with the characteristics of software development. These suggested measures have certain reference significance for medical device software development and quality control personnel, and technical reviewer and inspectors.

Combining Error Guessing and Logical Reasoning for Software Fault Localization via Deep Learning

Automated fault localization has been extensively studied to improve the effectiveness of software debugging. Existing automated fault localization methods neglect the guidance of the simple and easily available debugging information on fault localization. To bridge manual fault localization with automated fault localization, we propose a fault localization approach combining error guessing and logical reasoning via deep learning. The proposed approach simulates the actual debugging process. Specifically, developers’ debugging experience and context dependencies between methods are mapped into two different types of coverage matrices. The constructed matrices are fed to a convolutional neural network (CNN) to predict whether a method is buggy or not. To validate the effectiveness of the proposed approach, we designed and constructed the empirical study on the widely used Defect4J datasets. With respect to the top-n ([Formula: see text]) metric, our approach outperforms the state-of-the-art DeepFL and other five methods including Ochai, Muse, MULTRIC, TraPT and FLUCSS. Particularly, compared with the above methods, our approach has an improvement of 5–182% for top-1. In terms of MFR and MAR, the proposed approach is slightly lower than the best DeepFL but better than the other five methods. The approach we presented achieving the unification of manual and automatic debugging can aid in the improvement of fault localization accuracy.

A Software Reliability Model Considering a Scale Parameter of the Uncertainty and a New Criterion

It is becoming increasingly common for software to operate in various environments. However, even if the software performs well in the test phase, uncertain operating environments may cause new software failures. Traditional proposed software reliability models under uncertain operating environments suffer from the problem of being well-suited to special cases due to the large number of assumptions involved. To improve these problems, this study proposes a new software reliability model that assumes an uncertain operating environment. The new software reliability model is a model that minimizes assumptions and minimizes the number of parameters that make up the model, so that the model can be applied to general situations better than the traditional proposed software reliability models. In addition, various criteria based on the difference between the predicted and estimated values have been used in the past to demonstrate the superiority of the software reliability models. Also, we propose a new multi-criteria decision method that can simultaneously consider multiple goodness-of-fit criteria. The multi-criteria decision method using ranking is useful for comprehensive evaluation because it does not rely on individual criteria alone by ranking and weighting multiple criteria for the model. Based on this, 21 existing models are compared with the proposed model using two datasets, and the proposed model is found to be superior for both datasets using 15 criteria and the multi-criteria decision method using ranking.

#2519 Technical complications assessment of the novel SILENCIA© APD cycler

Abstract Background and Aims A new gravimetric peritoneal dialysis (PD) cycler, SILENCIA (Fresenius Medical Care, Bad Homburg, Germany), has been reported to achieve not only an adequate dialysis dose, but it is proposed also to be easy to use, cost-effective, and silent. In this study, we explored the alarms, cautions and number of terminated treatments due to potential errors and malfunctions of this new cycler. Method Over a period of 6 months, technical related device data was collected remotely, via a cloud-based infrastructure, including a Gateway and a Card Reader. The technical related data was then analyzed and visualized in a Dashboard. In special occasions, e.g., in case of device issues additional further technical data was collected on-site at patients’ home by a Service Technician. Results In total, 1860 technical related device data sets were collected in 3 different countries (Chile, Ecuador, and Turkey) from the treatments of 18 PD patients. 15 out of 32 patients were between 31 and 50 years old, and the mean PD duration was 2.14 years (n = 32). In total 9 alarms (5 per 1000 treatments) occurred, 5 being related to software bugs, 4 undetermined. 99.5% of the treatments were performed without alarms. In average, 5.3 cautions per 1000 treatments were observed, 94.7% due to blocked tubing system. Power failures only occurred in 4% of the treatments, with an average duration of 38.6 minutes. During power failures, the device does not wake up patients, proceeds as the power is back and treatment stays within schedule due to FlexPoint. A detailed list of alarms, cautions and power failures is presented in Table 1. Conclusion The results of this study demonstrate that SILENCIA supports PD treatment with a low number of alarms and cautions, which may help to achieve a smooth therapy for the patients, especially considering that the treatments are mainly performed during the night.

#2507 Enhancing critical care nephrology: analyzing continuous renal replacement therapy with a citrate-based anticoagulation focus

Abstract Background and Aims Continuous renal replacement therapy (CRRT) is pivotal for managing acute kidney injury (AKI) unstable patients. However, questions persist regarding prescription and the role of anticoagulation. This study aims to examine CRRT practices, focus on citrate-based anticoagulation and clarify contraindications. Method This study comprises a retrospective analysis of a patient cohort requiring CRRT and admitted to the Postoperative Intensive Care Unit and Burn Unit. Inclusion criteria involved the need for CRRT and adult age, with exclusion criteria based on treatment duration (<48 hours), reducing the initial cohort from 256 to 176 patients. Clinical variables, including gender, age, underlying pathology, CRRT modality, and anticoagulation type, were characterized. Subsequently, the association between anticoagulation type and clot formation within 48 hours and in-hospital mortality was investigated. Furthermore, an analysis of citrate-based CRRT was conducted, evaluating laboratory values (Table 1) at the beginning and 48 hours using the Wilcoxon test, and their correlation with clot development and mortality. Additionally, we used a logistic regression model to explore the relationship between AST, ALT, and lactate values with citrate accumulation. Finally, a survival analysis was conducted. Results The sample consisted of 66.5% males and 33.5% females, with an average age of 59.48 years. Reasons for admission are depicted in Fig. 1. CRRT modalities were 20.5% HDVVC, 76.1% HDFVVC, and 3.4% HFVVC. Anticoagulation included 35.2% none, 28.4% heparin, 33.5% citrate, and 2.8% alprostadil. Filter clotting occurred in 29.1% withing 48 hours. Heparin was associated with a 5.4-fold increased risk of clotting compared to citrate. Mortality was 63.3%, with no anticoagulation type showing a significant association. Parameters at CRRT initiation and 48 hours for citrate-based CRRT patients and Wilcoxon test results are presented in Table 1. Only 5.1% had filter clotting <48 hours, and no baseline parameters were significantly associated with an increased clotting risk. There was one case of citrate intoxication due to a programming error, which was not included. 76.81% had CaT/CaI <2.1, and 7.25% had CaT/CaI >2.1, requiring technique adjustments. Aminotransferases and lactate values, often considered as contraindications for citrate, were analyzed: The average AST and ALT values were 188.53 IU/l and 108.08 IU/l, respectively. When AST levels exceeded 5 times the normal range and ALT levels exceeded 3 times the normal range, a statistically significant association with CaT/CaI >2.1 at the 48-hour mark of CRRT utilization was observed (p < 0.001 for both).The average lactic acid value was 1.74 mg/dl. Having a lactic acid level >4 was associated (p < 0.05) with a higher risk of developing CaT/CaI >2.1. 54.2% of the sample died. No baseline parameters were significantly associated with it. At 48 hours, higher bicarbonate levels and metabolic alkalosis were associated with increased mortality (p < 0.05). The median survival time was 11 days (±4.38). Conclusion This study supports the use of citrate-based anticoagulation in CRRT, as no cases of citrate intoxication (CaT/CaI >2.5) were found. Nonetheless, a significantly elevated AST (>5 times normal), ALT (>3 times normal), and lactate (>4 mg/dl), were associated with a higher risk of citrate accumulation (CaT/CaI >2.1). Customized approaches and preventive measures to mitigate the development of metabolic alkalosis are essential for optimizing patient outcomes.

Empowering software reliability: Leveraging efficient fault detection and removal efficiency

Advancements in science and technology have led to the widespread use of computer systems in various applications, emphasizing the importance of software reliability. Software failures can have severe consequences, making thorough testing crucial. Software reliability growth models (SRGMs) play a significant role in enhancing reliability by predicting improvement over time. This article introduces a comprehensive approach to software reliability that incorporates a dynamic fault detection rate, along with fault removal efficiency. The fault detection rate measures the rate at which faults are identified during testing, reflecting the effectiveness of the testing process. By incorporating this dynamic component, the model provides a more accurate estimation of software reliability and enables adaptive testing strategies and resource allocation. Achieving a high fault detection rate is desirable, but organizations must consider the cost implications and strike a balance between reliability and time-to-market constraints. This article extends the analysis to calculate the optimal release time and optimal warranty period that minimize development costs, subject to the desired reliability. By considering these factors, development teams can make informed decisions regarding the timing of software release and the duration of the warranty period, optimizing both reliability and cost.

Online boxplot derived outlier detection

AbstractOutlier detection is a widely used technique for identifying anomalous or exceptional events across various contexts. It has proven to be valuable in applications like fault detection, fraud detection, and real-time monitoring systems. Detecting outliers in real time is crucial in several industries, such as financial fraud detection and quality control in manufacturing processes. In the context of big data, the amount of data generated is enormous, and traditional batch mode methods are not practical since the entire dataset is not available. The limited computational resources further compound this issue. Boxplot is a widely used batch mode algorithm for outlier detection that involves several derivations. However, the lack of an incremental closed form for statistical calculations during boxplot construction poses considerable challenges for its application within the realm of big data. We propose an incremental/online version of the boxplot algorithm to address these challenges. Our proposed algorithm is based on an approximation approach that involves numerical integration of the histogram and calculation of the cumulative distribution function. This approach is independent of the dataset’s distribution, making it effective for all types of distributions, whether skewed or not. To assess the efficacy of the proposed algorithm, we conducted tests using simulated datasets featuring varying degrees of skewness. Additionally, we applied the algorithm to a real-world dataset concerning software fault detection, which posed a considerable challenge. The experimental results underscored the robust performance of our proposed algorithm, highlighting its efficacy comparable to batch mode methods that access the entire dataset. Our online boxplot method, leveraging dataset distribution to define whiskers, consistently achieved exceptional outlier detection results. Notably, our algorithm demonstrated computational efficiency, maintaining constant memory usage with minimal hyperparameter tuning.