In order to monitor the operational status of the waste gas treatment system in a textile tentering and drying plant, two low-cost volatile organic compounds (VOC) sensors i.e. a metal oxide semiconductor sensor (MOS) and a photoionization detector sensor (PID), along with optical particulate matter (PM) sensors, were installed. Based on the measurement data, the stability and reliability of the monitoring system as well as the treatment effectiveness of the waste gas treatment system were assessed. The results showed that the water scrubber exhibited varying removal capacities for waste gases from the tentering and drying processes. While the electrostatic fume collector demonstrated excellent PM retention, it often resulted in increased VOC sensors readings at the discharge outlets. The correlation between VOC sensor readings was relatively strong across both high and low concentration levels, making it feasible to use the more affordable MOS sensors in place of the more expensive PID sensors. Overall, the data from the MOS sensors also provided some indication of PM concentrations. This study demonstrates that continuous monitoring of a waste gas treatment system using low-cost sensors can deliver valuable measurement data and serve as a control parameter for the operational management. This monitoring approach is particularly suitable for continuous monitoring of waste gas treatment systems in small and medium-sized enterprises.

In the dynamic world of finance, the application of Artificial Intelligence (AI) in pair trading strategies is gaining significant interest among scholars. Current AI research largely concentrates on regression analyses of prices or spreads between paired assets for formulating trading strategies. However, AI models typically exhibit less precision in regression tasks compared to classification tasks, presenting a challenge in refining the accuracy of pair trading strategies. In pursuit of high-performance labels to elevate the precision of classification models, this study advanced the Triple Barrier Labeling Method for enhanced compatibility with pair trading strategies. This refinement enables the creation of diverse label sets, each tailored to distinct barrier configurations. Focusing on achieving maximal profit or minimizing the Maximum Drawdown (MDD), Genetic Algorithms (GAs) were employed for the optimization of these labels. After optimization, the labels were classified into two distinct types: High Risk and High Profit (HRHP) and Low Risk and Low Profit (LRLP). These labels then serve as the foundation for training machine learning models, which are designed to predict future trading activities in the cryptocurrency market. Our approach, employing cryptocurrency price data from 9 November 2017 to 31 August 2022 for training and 1 September 2022 to 1 December 2023 for testing, demonstrates a substantial improvement over traditional pair trading strategies. In particular, models trained with HRHP signals realized a 51.42% surge in profitability, while those trained with LRLP signals significantly mitigated risk, marked by a 73.24% reduction in the MDD. This innovative method marks a significant advancement in cryptocurrency pair trading strategies, offering traders a powerful and refined tool for optimizing their trading decisions.

Due to a high concentration of particulate matter (PM10), the Korean Peninsula experienced its first poor air quality event of the year between November 19 and November 26, 2021. This study analyzes the reasons behind the occurrence of high-concentration PM10, using the real-time PM10 fugitive emission fluxes and meteorological data measured at two landfills for fly ash of coal-fired power plants located on the west coast. The real-time fugitive emission fluxes of PM10 were estimated at two different locations by a flux-gradient technique based on the eddy covariance method. The measurement results show a weak correlation between PM10 and various meteorological factors in the two places when PM10 levels are low. However, high PM10 concentrations were found to be strongly associated with the relative humidity of site A and the friction velocity of site B, respectively. High emission fluxes were observed at both sites under elevated temperature, high humidity, low wind speed, low frictional velocity, and atmospheric instability. The variation in weather patterns witnessed during periods of high PM10 concentrations in the two locations indicates that the causes of PM10 accumulation are different. The study demonstrates that the gradient-flux method's real-time measurement of fugitive emissions can explain the origin of high PM10 levels and provide essential data to efficiently regulate PM10.

In the software testing phase, software reliability growth models (SRGMs) are commonly used to evaluate the reliability of software systems. Traditional SRGMs are restricted by their assumption of a continuous growth pattern for the failure detection rate (FDR) throughout the testing phase. However, the assumption is compromised by Change-Point phenomena, where FDR fluctuations stem from variations in testing personnel or procedural modifications, leading to reduced prediction accuracy and compromised software reliability assessments. Therefore, the objective of this study is to improve software reliability prediction using a novel approach that combines genetic algorithm (GA) and deep learning-based SRGMs to account for the Change-point phenomenon. The proposed approach uses a GA to dynamically combine activation functions from various deep learning-based SRGMs into a new mutated SRGM called MuSRGM. The MuSRGM captures the advantages of both concave and S-shaped SRGMs and is better suited to capture the change-point phenomenon during testing and more accurately reflect actual testing situations. Additionally, failure data is treated as a time series and analyzed using a combination of Long Short -Term Memory (LSTM) and Attention mechanisms. To assess the performance of MuSRGM, we conducted experiments on three distinct failure datasets. The results indicate that MuSRGM outperformed the baseline method, exhibiting low prediction error (MSE) on all three datasets. Furthermore, MuSRGM demonstrated remarkable generalization ability on these datasets, remaining unaffected by uneven data distribution. Therefore, MuSRGM represents a highly promising advanced solution that can provide increased accuracy and applicability for software reliability assessment during the testing phase.

Edge computing is applied to various applications and is typically applied to autonomous driving software. As the self-driving system becomes complicated and the proportion of software increases, accidents caused by software defects increase. Just-in-time (JIT) defect prediction is a technique that identifies defects during the software development phase, which helps developers prioritize code inspection. Many researchers have proposed various JIT models, but it is difficult to find a case in which JIT defect prediction was performed on edge computing applications. In particular, due to the characteristic of self-driving software, which is frequently updated, there is a high risk of inducing defects into the update process. In this work, we propose a JIT defect prediction model via deep learning for edge computing applications called JIT4EA. Our research goal is to develop an effective model to predict defects in edge computing applications. To do this, we perform defect prediction on self-driving software, a representative edge computing application. We use pre-trained unified cross-modal pre-training for code representation (UniXCoder) to embed commit messages and code changes. We use bidirectional-LSTM(Bi-LSTM) for context and semantic learning. As a result of the experiment, it was confirmed that the proposed JIT4EA performed better than state-of-the-art methods and could reduce the code inspection effort.

Recently, Python is the most-widely used language in artificial intelligence (AI) projects requiring huge amount of CPU and memory resources, and long execution time for training. For saving the project duration and making AI software systems more reliable, it is inevitable to handle exceptions appropriately at the code level. However, handling exceptions highly relies on developer's experience. This is because, as an interpreter-based programming language, it does not force a developer to catch exceptions during development. In order to resolve this issue, we propose an approach to suggesting appropriate exceptions for the AI code segments during development after training exceptions from the existing handling statements in the AI projects. This approach learns the appropriate token units for the exception code and pretrains the embedding model to capture the semantic features of the code. Additionally, the attention mechanism learns to catch the salient features of the exception code. For evaluating our approach, we collected 32,771 AI projects using two popular AI frameworks (i.e. Pytorch and Tensorflow) and we obtained the 0.94 of Area under the Precision-Recall Curve (AUPRC) on average. Experimental results show that the proposed method can support the developer's exception handling with better exception proposal performance than the compared models.

소프트웨어 결함 예측은 효과적인 테스팅 자원 할당에 유용한 기법이다. 소프트웨어 교차 버전 결함 예측은 소프트웨어가 연속적인 버전으로 개발되는 환경을 반영하며, 소프트웨어 모듈은 버전 업데이트 과정을 거치며 추가되거나 삭제된다. 이러한 과정의 반복은 버전 간의 데이터 분포 차이를 발생시키는 원인이 되며, 데이터 분포 차이는 예측 성능을 저하시킨다. 딥 도메인 적응 기법은 컴퓨터 비전 분야에서 소스와 타겟 데이터의 분포 차이를 줄이는데 사용되는 기법이다. 본 연구의 목표는 다양한 적대적 학습 기반의 딥 도메인 적응 기법을 활용해 버전 간의 데이터 분포 차이를 줄이고, 결함 예측 성능이 가장 우수한 기법을 식별하는 것이다. 우리는 딥 도메인 적응 기법인 Domain-Adversarial Neural Network(DANN)와 Adversarial Discriminator Domain Apaptation(ADDA), Wasserstein Distance Guided Representation Learning(WDGRL)의 성능을 비교 실험하며, 소스 데이터의 구성에 따른 성능 차이를 확인한다. 또한 학습 과정에 사용되는 타겟 데이터의 비율에 따른 성능 차이와 세 가지 딥 도메인 적응 기법 중 결함 예측성능이 가장 우수한 DANN 모델의 하이퍼파라미터 설정에 따른 성능 차이를 확인한다. 실험 결과, DANN 모델은 타겟 버전을 제외한 이전 버전의 모든 데이터를 소스로 사용할 때 성능이 가장 우수하다. 특히 DANN 모델의 은닉층 수를 3개로 설정할 때 가장 우수한 성능을 보였다. 또한 딥 도메인 적응 기법을 적용할 때, 학습 과정에 사용되는 타겟 데이터가 많을수록 우수한 성능을 보인다. 본 연구를 통해 향후 다양한 딥 도메인 적응 기법이 소프트웨어 교차 버전 결함 예측에 활용될 수 있을 것으로 기대한다.

With recent increases in the number of network-connected devices, the number of edge computing services that provide similar functions has increased. Therefore, it is important to recommend an optimal edge computing service, based on quality-of-service (QoS). However, in the real world, there is a cold-start problem in QoS data: highly sparse invocation. Therefore, it is difficult to recommend a suitable service to the user. Deep learning techniques were applied to address this problem, or context information was used to extract deep features between users and services. However, edge computing environment has not been considered in previous studies. Our goal is to predict the QoS values in real edge computing environments with improved accuracy. To this end, we propose a GAIN-QoS technique. It clusters services based on their location information, calculates the distance between services and users in each cluster, and brings the QoS values of users within a certain distance. We apply a Generative Adversarial Imputation Nets (GAIN) model and perform QoS prediction based on this reconstructed user service invocation matrix. When the density is low, GAIN-QoS shows superior performance to other techniques. In addition, the distance between the service and user slightly affects performance. Thus, compared to other methods, the proposed method can significantly improve the accuracy of QoS prediction for edge computing, which suffers from cold-start problem.

In this study, the formation mechanism of deformation bands (DBs) observed in AA1100 after plane forging (PF) and return-plane forging (R-PF) during multi-axis diagonal forging (MADF) was unraveled via microtexture analysis using a rate-sensitive model and crystal plasticity finite element method (CPFEM). Blurry bands with an orientation relationship of mainly LAGBs were developed in the deformed grains in the PF specimen. The heterogeneity of the orientation stability (Q) that occurred within the blurry bands and their surrounding regions contributed to the development of the DBs in the deformed grains in the R-PF specimen. Microstructure-based CPFEM successfully simulated the development of DBs within deformed grains during R-PF. CPFEM revealed that heterogeneity of the crystallographic orientation that occurred within the blurry bands and their surrounding regions inside the PF specimen was the main cause of the formation of DBs in the deformed grains during R-PF.

In Version Control System (VCS), a developer frequently uploads multiple tasks such as adding features, code refactoring, and fixing bugs, into a single commit and crumbles each task's summary when writing a commit message. It causes code readers to feel challenged in understanding the developer's past tasks within the commit history. To resolve this issue, we propose an automatic approach to generating a task summary to help comprehend multiple mixed tasks in a commit and developed tool support named Task summary Generator (TsGen). In our approach, we use the commit with a single task as input and identify the task to sort its elements sequentially. Then we generate feature vectors from each sorted element to train the Neural Machine Translation (NMT) model. Based on the trained NMT model, we generate the feature vector from each task of a commit with multiple tasks and put each of them into the model to provide the task summary. In evaluation, we compared the performance of TsGen with two existing methods for nine open-source projects. As a result, TsGen outperformed CoDiSum and Jiang's NMT by 52.08% and 28.07% in BiLingual Evaluation Understudy (BLEU) scores. In addition, the human evaluation was carried out to demonstrate that TsGen helps understand mixed tasks in a commit and gained a 0.27 higher preference than the actual commit message.

As a future game-changer in various industries, cryptocurrency is attracting people’s attention. Cryptocurrency is issued on blockchain and managed through a blockchain wallet application. The blockchain wallet manages user’s digital assets and authenticates a blockchain user by checking the possession of a user’s private key. The mnemonic code technique represents the most widely used method of generating and recovering a private key in blockchain wallet applications. However, the mnemonic code technique does not consider usability to generate and recover a user’s private key. In this study, we propose a novel approach for private key generation and recovery. Our approach is based on the idea that a user can hold long-term memory from distinctive pictures. The user can generate a private key by providing pictures and the location of the pictures. For recovering a private key, the user identifies the locations of the pictures that are used in the private key generation process. In this paper, we experiment with the security and usability of our approach and confirm that our proposed approach is sufficiently secure compared to the mnemonic code technique and accounts for usability.

A software defect prediction (SDP) model identifies the defect-prone modules. Setting appropriate parameters in an SDP model is critical because it affects the model performance. In a recent study, parameters were automatically explored using an optimization algorithm. However, such studies did not explore all the parameters that could be handled in the SDP process from preprocessing to model building, but only optimized parameters in some modeling process steps, such as feature selection or model building. Our goal is to improve the model performance by optimizing parameters across the entire SDP process. For this, we propose a cost-sensitive decision tree based on harmony search (HS-CSDT). HS-CSDT uses a harmony search algorithm to simultaneously identify the optimal feature set, regularization technique, class weight, and decision tree hyperparameters. We compared HS-CSDT against the methods in related studies in terms of probability of detection, probability of false alarm, G-measure, and file inspection reduction in the evaluation of 28 open-source projects. The results of the effect size using Cohen's d reveal that HS-CSDT provides a statistically better performance than methods in related work. Experimental results show that optimizing the identified parameters throughout the entire SDP modeling process by using the optimization algorithm helps improve the model performance. In summary, HS-CSDT shows excellent defect prediction performance by automatically allocating an appropriate parameter set according to the software project. Thus, the model can help effectively allocate limited quality assurance resources.

In this study, the deformation bands (DBs) in AA1100 after plane forging (PF) and return-plane forging (R-PF) during multi-axis diagonal forging (MADF) were observed using microtexture analysis. Typical texture components (plane-strain texture components) observed in face-centered cubic (FCC) materials deformed by plane-strain compression were developed in the PF specimen. Blurry bands with an orientation relationship of mainly LAGBs were developed in the deformed grains in the PF specimen. Slip trace analysis and Schmid factor (SF) calculations revealed the correlation between the blurry bands and the slip behavior. Regions of matrix and DBs inside the deformed grains of the R-PF specimen were distinguished mostly by their HAGBs. The heterogeneity of the Taylor factor (M) values that occurred within the blurry bands and their surrounding regions induced the development of the DBs in the R-PF specimen. The heterogeneities of the deformation behavior between the inside and the outside of the blurry bands that developed in the grains of the PF specimen during the R-PF were a direct reason for the generation of DBs in the deformed grains of the R-PF specimen. The transmission-EBSD (t-EBSD) technique revealed that the average misorientation angle inside the DBs was higher than that inside the matrix.

Gradle is one of the widely used tools to automatically build a software project. While developers execute the Gradle build for projects, they face various build errors in practice. However, fixing build errors is not easy because developers should manually find out the cause of the build error and its resolution on their project. For this reason, developers spend much time fixing them, and especially it can be worse if a developer lacks the experience of handling build errors. To address this issue, we propose a novel approach named Gradle-AutoFix to automatically fix build errors along with providing their causes and resolutions. In this approach, we collect build errors to group their causes and resolutions and then generate feature vectors from build error messages by applying Bag-of-Word (BoW), Term Frequency-Inverse Document Frequency (TF-IDF), Bigram, and an embedding layer. The feature vectors are utilized for training two classification models on cause and resolution. Next, we analyze fixing patterns and define seven resolution rules to fix the build error automatically. Based on our trained models and defined resolution rules, we built Gradle-AutoFix. For the evaluation, we measured how appropriately Gradle-AutoFix provides causes of build errors and resolutions. As a result, we obtained 96% and 91% accuracy, respectively. Also, we assessed how properly Gradle-AutoFix fixes the project’s build error based on the seven resolution rules. The outcome showed a 64.5% build error resolution rate for 231 projects.

소프트웨어 결함 분석은 소프트웨어의 품질을 보증하는 데에 중요한 연구 분야 중 하나이다. 결함 분석은 예측하고자 하는 출력에 따라 결함 예측, 결함 회귀, 결함 심각도 예측 등으로 분류된다. 각 출력에서 뛰어난 예측성능을 보이기 위해 많은 연구가 제안되었다. 하지만, 실제 사용자가 여러 모델로 여러 출력 정보를 얻고자 하는 경우, 모델들의 출력 간 일관성이 낮은 문제가 존재한다. 본 논문에서는 결함 · 비결함, 버그의 수, 버그의 심각도 3가지 출력을 동시에 하나의 모델로 예측해 출력 간 일관성을 높이고자 한다. 이를 위해 XGBoost와 다중 출력 1차원 합성곱 신경망을 결합한 모델인 X-MOCNN(XGBoost with Multi-Output 1-d Convolutional Neural Network)을 제안한다. 이 모델의 효용성을 검증하기 위해 단일 출력 신경망 모델, 머신 러닝 모델과 성능을 비교하였다. 그 결과 X-MOCNN은 기존의 단일 출력신경망 모델 대비 높은 성능을 보였으며 각 예측 결과 간 일관성 정도도 뛰어났다. 이를 통해 소프트웨어 프로젝트에서 하나의 모델로 여러 출력을 예측해 품질보증 자원을 효과적으로 할당할 수 있다.

소프트웨어 결함 예측은 결함이 야기될 모듈을 식별해 한정된 품질 보증 자원을 효과적으로 배분하는데 도움을 준다. 소프트웨어 결함 데이터는 비결함 인스턴스의 수가 결함 인스턴스의 수보다 많은 클래스 불균형 문제를 겪는다. 대부분의 기계 학습에서 특정 클래스의 인스턴스 비율이 한쪽으로 치우치게 되면 결함 예측 성능에 부정적인 영향을 끼친다. 따라서 본 연구에서는 생성적 적대 신경망 모델(Generative Adversarial Network, GAN)을 사용해 클래스 불균형 문제를 해결하고, 결함 예측 성능 향상을 목표로 한다. 이를 위해, 본 연구에서는 여러 종류의 GAN 모델 중 소프트웨어 결함 예측에 적합한 모델은 무엇인지 비교하고, 관련 연구에서 적용하지 않았던 GAN 모델들의 적용성 여부를 확인한다. 본 연구에서는 이미지 생성에 최적화되어 있는 Vanilla-GAN(GAN)과 Conditional GAN(cGAN), Wasserstein GAN(WGAN) 모델을 소프트웨어 결함 예측 데이터에 적합하게 개조한 후, 개조한 GAN과 cGAN, WGAN, Tabular GAN(TGAN), Modeling Tabular data using Conditional GAN(CTGAN)의 성능을 비교 실험한다. 실험 결과, CTGAN 모델이 소프트웨어 결함 예측 데이터에 적합함을 보인다. 또한 CTGAN의 하이퍼파라미터 중 결함 발견율(Recall)을 높이고, 결함 오보율(Probability of False Alarm, PF)를 낮추는 하이퍼파라미터 값은 무엇인지 민감도 분석을 수행한다. 실험 결과, 데이터셋에 따라 하이퍼파라미터를 조정해야 함을 보였다. 우리의 제안한 기법이 소프트웨어 결함 예측의 성능을 향상시켜 한정된 자원을 효과적으로 할당하는데 도움이 될 것이라고 기대한다.

소프트웨어 실패의 주요 원인들 중 하나로 컨피규레이션 버그가 있다. 소프트웨어 조직들은 이슈 트래킹 시스템을 통해 버그 리포트들을 수집하고 관리하는데, 버그 할당자는 해당 버그가 컨피규레이션 버그인지 식별하는데 시간을 소비할 수 있다. 컨피규레이션 버그를 예측하는 방법을 통해 버그 할당자의 의사 결정에 도움을 줘 노력을 줄일 수 있다. 본 논문에서는 텍스트 마이닝 기법과 차원 축소 기법을 이용하여 향상된 분류 모델을 제안한다. 본 논문은 6개의 오픈 소스 소프트웨어 프로젝트로부터 4,457개의 버그 리포트를 추출하고 컨피규레이션 버그 리포트를 분류하는 모델을 학습하고 예측 성능을 평가한다. 가장 좋은 성능을 보이는 방법은 Bag of Words로 피쳐를 추출하고 선형판별분석(LDA: Linear Discriminant Analysis)를 이용하여 피쳐의 차원을 축소 후 SMOTEENN 샘플링 기법을 이용하여 k-Nearest Neighbors 모델을 사용한다. 이에 대한 AUC 값은 0.9812이고 MCC가 0.942이다. 이는 Xia et al.의 방법보다 더 좋은 성능을 보이며, 이전 연구에서의 클래스 불균형 문제를 해결한다. 이러한 향상된 컨피규레이션 버그 리포트 예측을 통해, 이를 버그 할당자의 의사 결정에 필요한 정보를 줄 수 있거나 시간을 단축시킬 수 있다.