Terms Of F-score Research Articles

IGROWL: Improved Group Detection With Link Prediction

One of the main challenges robots need to overcome is crowd analysis. Crowd analysis deals with the detection of individuals and interaction groups as well as the recognition of their activities. This paper focuses on the detection of conversational groups, where there have been a number of approaches addressing this problem in both supervised and unsupervised ways. Supervised bottom-up approaches primarily relied on pairwise affinity matrices and were limited to static, third-person views. In this work, we present our approach based on Graph Neural Networks (GNNs) to the problem of interaction group detection, called improved Group Detection With Link Prediction (iGROWL). iGROWL utilises the fact that interaction groups exist in certain inherent spatial configurations and improves its predecessor, GROWL, by introducing an ensemble learning-based sample balancing technique to the algorithm. Our results show that iGROWL outperforms other state-of-the-art methods by 16.7% and 26.4% in terms of F1-score when evaluated on the Salsa Poster Session and Cocktail Party datasets, respectively. Moreover, we show that sample balancing with GNNs is not trivial, but consistent results can be achieved by employing ensemble learning.

Semi-Supervised Semantic Role Labeling with Bidirectional Language Models

The recent success of neural networks in NLP applications has provided a strong impetus to develop supervised models for semantic role labeling (SRL) that forego the requirement for extensive feature engineering. Recent state-of-the-art approaches require high-quality annotated datasets that are costly to obtain and almost unavailable for low-resource languages. We present a semi-supervised approach that utilizes both labeled and unlabeled data to provide performance improvement over a mere supervised SRL model. We show that our proposed semi-supervised SRL model provides larger improvement over a supervised model in the scenario where labeled training data size is small. Our SRL system leverages unlabeled data under the language modeling paradigm. We demonstrate that the incorporation of a self pre-trained bidirectional language model (S-PrLM) into a SRL system can help in SRL performance improvement by learning composition functions from the unlabeled data. Previous researches have concluded that syntax information is very useful for high-performing SRL systems, so we incorporate syntax information by employing an unsupervised approach to leverage dependency path information to connect argument candidates in vector space, which helps in distinguishing arguments with similar contexts but different syntactic functions. The basic idea is to connect predicate ( w p ) with argument candidate ( w a ) with the dependency path ( r ) between them in the embedding space. Experiments on the CoNLL-2008 and CoNLL-2009 datasets confirm that our full SRL model outperforms previous best models in terms of F 1 score.

Weakly Supervised U-Net with Limited Upsampling for Sound Event Detection

Sound event detection (SED) is the task of finding the identities of sound events, as well as their onset and offset timings from audio recordings. When complete timing information is not available in the training data, but only the event identities are known, SED should be solved by weakly supervised learning. The conventional U-Net with global weighted rank pooling (GWRP) has shown a decent performance, but extensive computation is demanded. We propose a novel U-Net with limited upsampling (LUU-Net) and global threshold average pooling (GTAP) to reduce the model size, as well as the computational overhead. The expansion along the frequency axis in the U-Net decoder was minimized, so that the output map sizes were reduced by 40% at the convolutional layers and 12.5% at the fully connected layers without SED performance degradation. The experimental results on a mixed dataset of DCASE 2018 Tasks 1 and 2 showed that our limited upsampling U-Net (LUU-Net) with GTAP was about 23% faster in training and achieved 0.644 in audio tagging and 0.531 in weakly supervised SED tasks in terms of F1 scores, while U-Net with GWRP showed 0.629 and 0.492, respectively. The major contribution of the proposed LUU-Net is the reduction in the computation time with the SED performance being maintained or improved. The other proposed method, GTAP, further improved the training time reduction and provides versatility for various audio mixing conditions by adjusting a single hyperparameter.

Personalized and motion-based human activity recognition with transfer learning and compressed deep learning models

Human activity recognition (HAR) enables the recognition of the activities of daily living using signals from motion sensors integrated into mobile and wearable devices. One of the challenges is the uniqueness of each individual with his/her different characteristics. A general model trained without user data may perform poorly on specific users. Another challenge is running deep learning (DL) models on mobile and wearable devices due to their limited resources. In this paper, to cope with these two challenges, we use transfer learning to build personalized models and model compression for running DL algorithms. We examine the impact of different DL architectures, the number of layers to be fine-tuned, the amount of user training data, and the transfer to new datasets on the performance of HAR. We compare the performance of the transferred models with general and user-specific models in terms of F1 score, training and inference time.

Using Federated Learning for Collaborative Intrusion Detection Systems

Neural networks have become cutting edge machine learning models for detecting network attacks. Traditional implementations provide fast and accurate predictions, but require centralised storage of labelled historical data for training. This solution is not always suitable for real-world applications, where regulatory constraints and privacy concerns hamper the collection of sensitive data into a single server. Federated Learning has recently been proposed as a framework for training a centralised model without the need to share data between different providers. We use the CICIDS2017 dataset provided by the Canadian Institute of Cybersecurity to demonstrate the benefits of Neural Networks-based Federated Learning for the detection of the most relevant types of network attacks. We conclude that a federated-trained neural network outperforms locally-trained models (at isoarchitecture) in terms of F1-score and False Negative detection ratio. Further, such model has a minor loss of performance and convergence rapidity compared to a model trained over a hypothetical centralised dataset.

Fatigue Detection System in Construction Site Using Extension Based Equilibrium with Capsule Autoencoder Network

Fatigue detection of workers is an important factor in construction site monitoring. Nowadays, worker exhaustion on construction sites causes tiredness and drowsiness. The prediction of mental exhaustion is critical because the job has increased over the years. Accurate fatigue detection is important for analyzing the stress level of work on construction sites. However, recording worker activities and detecting fatigue is critical for site supervisors. Over the last century, there has been an increasing trend towards vision-based action recognition. The identification of worker activities in far-field surveillance video has received good attention. Henceforth, this research proposed a hierarchical statistical approach for detecting worker activity in far-field security footage. In this paper, extension based equilibrium with capsule autoencoder network is proposed for fatigue detection. To improve the prediction performance, video data is collected for monitoring. Initially, the video dataset is converted into frames, and these frames are pre-processed using normalization. Afterwards, statistical feature extraction techniques are extracted from the normalized frames to improve the prediction performance. It helps to identify the symptoms of fatigue detection. The persons with the symptoms are identified from the extraction. Feature selection is performed using hybrid Battle Royale Optimization and Particle Swarm Optimization algorithms (hybrid BRO-PSO). Finally, based on the selected features, fatigue classification is performed using extended equilibrium optimization with Capsule Autoencoder (EECAEN). The performance of the proposed scheme is compared with different existing approaches. The performance of fatigue detection is analyzed with real time video dataset. The performance of the proposed approach is compared with different existing techniques such as Convolution Neural Network (CNN), Long Short-Term Memory (LSTM), and Recurrent Neural Network (RNN). The proposed method outperforms the existing approaches in terms of F1-score, accuracy, recall, precision, specificity, and sensitivity.

Towards Predicting Length of Stay and Identification of Cohort Risk Factors Using Self-Attention-Based Transformers and Association Mining: COVID-19 as a Phenotype.

Predicting length of stay (LoS) and understanding its underlying factors is essential to minimizing the risk of hospital-acquired conditions, improving financial, operational, and clinical outcomes, and better managing future pandemics. The purpose of this study was to forecast patients' LoS using a deep learning model and to analyze cohorts of risk factors reducing or prolonging LoS. We employed various preprocessing techniques, SMOTE-N to balance data, and a TabTransformer model to forecast LoS. Finally, the Apriori algorithm was applied to analyze cohorts of risk factors influencing hospital LoS. The TabTransformer outperformed the base machine learning models in terms of F1 score (0.92), precision (0.83), recall (0.93), and accuracy (0.73) for the discharged dataset and F1 score (0.84), precision (0.75), recall (0.98), and accuracy (0.77) for the deceased dataset. The association mining algorithm was able to identify significant risk factors/indicators belonging to laboratory, X-ray, and clinical data, such as elevated LDH and D-dimer levels, lymphocyte count, and comorbidities such as hypertension and diabetes. It also reveals what treatments have reduced the symptoms of COVID-19 patients, leading to a reduction in LoS, particularly when no vaccines or medication, such as Paxlovid, were available.

Two simple methods to improve the accuracy of the genomic selection methodology

BackgroundGenomic selection (GS) is revolutionizing plant and animal breeding. However, still its practical implementation is challenging since it is affected by many factors that when they are not under control make this methodology not effective. Also, due to the fact that it is formulated as a regression problem in general has low sensitivity to select the best candidate individuals since a top percentage is selected according to a ranking of predicted breeding values.ResultsFor this reason, in this paper we propose two methods to improve the prediction accuracy of this methodology. One of the methods consist in reformulating the GS (nowadays formulated as a regression problem) methodology as a binary classification problem. The other consists only in a postprocessing step that adjust the threshold used for classification of the lines predicted in its original scale (continues scale) to guarantee similar sensitivity and specificity. The postprocessing method is applied for the resulting predictions after obtaining the predictions using the conventional regression model. Both methods assume that we defined with anticipation a threshold, to divide the training data as top lines and not top lines, and this threshold can be decided in terms of a quantile (for example 80%, 90%, etc.) or as the average (or maximum) of the performance of the checks.In the reformulation method it is required to label as one those lines in the training set that are equal or larger than the specified threshold and as zero otherwise. Then we train a binary classification model with the conventional inputs, but using the binary response variable in place of the continuous response variable. The training of the binary classification should be done to guarantee a more similar sensitivity and specificity, to guarantee a reasonable probability of classification of the top lines.ConclusionsWe evaluated the proposed models in seven data sets and we found that the two proposed methods outperformed by large margin the conventional regression model (by 402.9% in terms of sensitivity, by 110.04% in terms of F1 score and by 70.96% in terms of Kappa coefficient, with the postprocessing methods). However, between the two proposed methods the postprocessing method was better than the reformulation as binary classification model. The simple postprocessing method to improve the accuracy of the conventional genomic regression models avoid the need to reformulate the conventional regression models as binary classification models with similar or better performance, that significantly improve the selection of the top best candidate lines. In general both proposed methods are simple and can easily be adopted for use in practical breeding programs, with the guarantee that will improve significantly the selection of the top best candidates lines.

A self-inspected adaptive SMOTE algorithm (SASMOTE) for highly imbalanced data classification in healthcare

In many healthcare applications, datasets for classification may be highly imbalanced due to the rare occurrence of target events such as disease onset. The SMOTE (Synthetic Minority Over-sampling Technique) algorithm has been developed as an effective resampling method for imbalanced data classification by oversampling samples from the minority class. However, samples generated by SMOTE may be ambiguous, low-quality and non-separable with the majority class. To enhance the quality of generated samples, we proposed a novel self-inspected adaptive SMOTE (SASMOTE) model that leverages an adaptive nearest neighborhood selection algorithm to identify the “visible” nearest neighbors, which are used to generate samples likely to fall into the minority class. To further enhance the quality of the generated samples, an uncertainty elimination via self-inspection approach is introduced in the proposed SASMOTE model. Its objective is to filter out the generated samples that are highly uncertain and inseparable with the majority class. The effectiveness of the proposed algorithm is compared with existing SMOTE-based algorithms and demonstrated through two real-world case studies in healthcare, including risk gene discovery and fatal congenital heart disease prediction. By generating the higher quality synthetic samples, the proposed algorithm is able to help achieve better prediction performance (in terms of F1 score) on average compared to the other methods, which is promising to enhance the usability of machine learning models on highly imbalanced healthcare data.

Optimizing smart contract vulnerability detection via multi-modality code and entropy embedding

Smart contracts have been widely used in the blockchain world these years, and simultaneously vulnerability detection has gained more and more attention due to the staggering economic losses caused by the attacker. Existing tools that analyze vulnerabilities for smart contracts heavily rely on rules predefined by experts, which are labour-intense and require domain knowledge. Moreover, predefined rules tend to be misconceptions and increase the risk of crafty potential back-doors in the future. Recently, researchers mainly used static and dynamic execution analysis to detect the vulnerabilities of smart contracts and have achieved acceptable results. However, the dynamic method cannot cover all the program inputs and execution paths, which leads to some vulnerabilities that are hard to detect. The static analysis method commonly includes symbolic execution and theorem proving, which requires using constraints to detect vulnerability. These shortcomings show that traditional methods are challenging to apply and expand on a large scale. This paper aims to detect vulnerabilities via the Bug Injection framework and transfer learning techniques. First, we train a Transformer encoder using multi-modality code, which contains source code, intermediate representation, and assembly code. The input code consists separately of Solidity source code, intermediate representation, and assembly code. Specifically, we translate source code into the intermediate representation and decompile the byte code into assembly code by the EVM compiler. Then, we propose a novel entropy embedding technique, which combines token embedding, segment embedding, and positional embedding of the Transformer encoder in our approach. After that, we utilize the Bug Injection framework to automatically generate specific types of buggy code for fine-tuning and evaluating the performance of vulnerability detection. The experimental results show that our proposed approach improves the performance in detecting reentrancy vulnerabilities and timestamp dependence. Moreover, our approach is more flexible and scalable than static and dynamic analysis approaches in detecting smart contract vulnerabilities. Our approach improves the baseline approaches by an average of 11.89% in term of F1 score.

Influence of Hyperparameters in Deep Learning Models for Coffee Rust Detection

Most of the world’s crops can be attacked by various diseases or pests, affecting their quality and productivity. In recent years, transfer learning with deep learning (DL) models has been used to detect diseases in maize, tomato, rice, and other crops. In the specific case of coffee, some recent works have used fixed hyperparameters to fine-tune the pre-trained models with the new dataset and/or applied data augmentation, such as image patching, to improve classifier performance. However, a detailed evaluation of the impact of architecture (e.g., backbone) and training (e.g., optimizer and learning rate) hyperparameters on the performance of coffee rust classification models has not been performed. Therefore, this paper presents a comprehensive study of the impact of five types of hyperparameters on the performance of coffee rust classification models. Specifically, eight pre-trained models are compared, each with four different amounts of transferred layers and three different numbers of neurons in the fully-connected (FC) layer, and the models are fine-tuned with three types of optimizers, each with three learning rate values. Comparing more than 800 models in terms of F1-score and accuracy, it is identified that the type of backbone is the hyperparameter with the greatest impact (with differences between models of up to 70%), followed by the optimizer (with differences of up to 20%). At the end of the study, specific recommendations are made on the values of the most suitable hyperparameters for the identification of this type of disease in coffee crops.

Towards complete tree crown delineation by instance segmentation with Mask R–CNN and DETR using UAV-based multispectral imagery and lidar data

Precise single tree delineation allows for a more reliable determination of essential parameters such as tree species, height and vitality. Methods of instance segmentation are powerful neural networks for detecting and segmenting single objects and have the potential to push the accuracy of tree segmentation methods to a new level. In this study, two instance segmentation methods, Mask R–CNN and DETR, were applied to precisely delineate single tree crowns using multispectral images and images generated from UAV lidar data. The study area was in Bavaria, 35 km north of Munich (Germany), comprising a mixed forest stand of around 7 ha characterised mainly by Norway spruce (Picea abies) and large groups of European beeches (Fagus sylvatica) with 181–236 trees per ha. The data set, consisting of multispectral images and lidar data, was acquired using a Micasense RedEdge-MX dual camera system and a Riegl miniVUX-1UAV lidar scanner, both mounted on a hexacopter (DJI Matrice 600 Pro). At an altitude of approximately 85 m, two flight missions were conducted at an airspeed of 5 m/s, leading to a ground resolution of 5 cm and a lidar point density of 560 points/m2. In total, 1408 trees were marked by visual interpretation of the remote sensing data for training and validating the classifiers. Additionally, 125 trees were surveyed by tacheometric means used to test the optimized neural networks. The evaluations showed that segmentation using only multispectral imagery performed slightly better than with images generated from lidar data. In terms of F1 score, Mask R–CNN with color infrared (CIR) images achieved 92% in coniferous, 85% in deciduous and 83% in mixed stands. Compared to the images generated by lidar data, these scores are the same for coniferous and slightly worse for deciduous and mixed plots, by 4% and 2%, respectively. DETR with CIR images achieved 90% in coniferous, 81% in deciduous and 84% in mixed stands. These scores were 2%, 1%, and 2% worse, respectively, compared to the lidar data images in the same test areas. Interestingly, four conventional segmentation methods performed significantly worse than CIR-based and lidar-based instance segmentations. Additionally, the results revealed that tree crowns were more accurately segmented by instance segmentation. All in all, the results highlight the practical potential of the two deep learning-based tree segmentation methods, especially in comparison to baseline methods.

Modeling Extractive Question Answering Using Encoder-Decoder Models with Constrained Decoding and Evaluation-Based Reinforcement Learning

Extractive Question Answering, also known as machine reading comprehension, can be used to evaluate how well a computer comprehends human language. It is a valuable topic with many applications, such as in chatbots and personal assistants. End-to-end neural-network-based models have achieved remarkable performance on these tasks. The most frequently used approach to extract answers with neural networks is to predict the answer’s start and end positions in the document, independently or jointly. In this paper, we propose another approach that considers all words in an answer jointly. We introduce an encoder-decoder model to learn from all words in the answer. This differs from previous works. which usually focused on the start and end and ignored the words in the middle. To help the encoder-decoder model to perform this task better, we employ evaluation-based reinforcement learning with different reward functions. The results of an experiment on the SQuAD dataset show that the proposed method can outperform the baseline in terms of F1 scores, offering another potential approach to solve the extractive QA task.

A Span-based Target-aware Relation Model for Frame-semantic Parsing

Frame-semantic Parsing (FSP) is a challenging and critical task in Natural Language Processing (NLP). Most of the existing studies decompose the FSP task into frame identification (FI) and frame semantic role labeling (FSRL) subtasks, and adopt a pipeline model architecture that clearly causes error propagation problem. However, recent jointly learning models aim to address the above problem and generally treat FSP as a span-level structured prediction task, which, unfortunately, leads to cascading error propagation problem between roles and less-efficient solutions due to huge search space of roles. To address these problems, we reformulate the FSRL task into a target-aware relation classification task and propose a novel and lightweight jointly learning framework that simultaneously processes three subtasks of FSP, including frame identification, argument identification, and role classification. The novel task formulation and jointly learning with interaction mechanisms among subtasks can help improve the overall system performance and reduce the search space and time complexity, compared with existing methods. Extensive experimental results demonstrate that our proposed model significantly outperforms 10 state-of-the-art models in terms of F1 score across two benchmark datasets.

Combining BERT with TCN-BiGRU for enhancing Arabic aspect category detection

Aspect-based sentiment analysis (ABSA) is a challenging task of sentiment analysis that aims at extracting the discussed aspects and identifying the sentiment corresponding to each aspect. We can distinguish three main ABSA tasks: aspect term extraction, aspect category detection (ACD), and aspect sentiment classification. Most Arabic ABSA research has relied on rule-based or machine learning-based methods, with little attention to deep learning techniques. Moreover, most existing Arabic deep learning models are initialized using context-free word embedding models, which cannot handle polysemy. Therefore, this paper aims at overcoming the limitations mentioned above by exploiting the contextualized embeddings from pre-trained language models, specifically the BERT model. Besides, we combine BERT with a temporal convolutional network and a bidirectional gated recurrent unit network in order to enhance the extracted semantic and contextual features. The evaluation results show that the proposed method has outperformed the baseline and other models by achieving an F1-score of 84.58% for the Arabic ACD task. Furthermore, a set of methods are examined to handle the class imbalance in the used dataset. Data augmentation based on back-translation has shown its effectiveness through enhancing the first results by an overall improvement of more than 3% in terms of F1-score.

INSnet: a method for detecting insertions based on deep learning network

BackgroundMany studies have shown that structural variations (SVs) strongly impact human disease. As a common type of SV, insertions are usually associated with genetic diseases. Therefore, accurately detecting insertions is of great significance. Although many methods for detecting insertions have been proposed, these methods often generate some errors and miss some variants. Hence, accurately detecting insertions remains a challenging task.ResultsIn this paper, we propose a method named INSnet to detect insertions using a deep learning network. First, INSnet divides the reference genome into continuous sub-regions and takes five features for each locus through alignments between long reads and the reference genome. Next, INSnet uses a depthwise separable convolutional network. The convolution operation extracts informative features through spatial information and channel information. INSnet uses two attention mechanisms, the convolutional block attention module (CBAM) and efficient channel attention (ECA) to extract key alignment features in each sub-region. In order to capture the relationship between adjacent subregions, INSnet uses a gated recurrent unit (GRU) network to further extract more important SV signatures. After predicting whether a sub-region contains an insertion through the previous steps, INSnet determines the precise site and length of the insertion. The source code is available from GitHub at https://github.com/eioyuou/INSnet.ConclusionExperimental results show that INSnet can achieve better performance than other methods in terms of F1 score on real datasets.

Detecting mild cognitive impairment and dementia in older adults using naturalistic driving data and interaction-based classification from influence score

Several recent studies indicate that atypical changes in driving behaviors appear to be early signs of mild cognitive impairment (MCI) and dementia. These studies, however, are limited by small sample sizes and short follow-up duration. This study aims to develop an interaction-based classification method building on a statistic named Influence Score (i.e., I-score) for prediction of MCI and dementia using naturalistic driving data collected from the Longitudinal Research on Aging Drivers (LongROAD) project. Naturalistic driving trajectories were collected through in-vehicle recording devices for up to 44 months from 2977 participants who were cognitively intact at the time of enrollment. These data were further processed and aggregated to generate 31 time-series driving variables. Because of high dimensional time-series features for driving variables, we used I-score for variable selection. I-score is a measure to evaluate variables’ ability to predict and is proven to be effective in differentiating between noisy and predictive variables in big data. It is introduced here to select influential variable modules or groups that account for compound interactions among explanatory variables. It is explainable regarding to what extent variables and their interactions contribute to the predictiveness of a classifier. In addition, I-score boosts the performance of classifiers over imbalanced datasets due to its association with the F1 score. Using predictive variables selected by I-score, interaction-based residual blocks are constructed over top I-score modules to generate predictors and ensemble learning aggregates these predictors to boost the prediction of the overall classifier. Experiments using naturalistic driving data show that our proposed classification method achieves the best accuracy (96%) for predicting MCI and dementia, followed by random forest (93%) and logistic regression (88%). In terms of F1 score and AUC, our proposed classifier achieves 98% and 87%, respectively, followed by random forest (with an F1 score of 96% and an AUC of 79%) and logistic regression (with an F1 score of 92% and an AUC of 77%). The results indicate that incorporating I-score into machine learning algorithms could considerably improve the model performance for predicting MCI and dementia in older drivers. We also performed the feature importance analysis and found that the right to left turn ratio and the number of hard braking events are the most important driving variables to predict MCI and dementia.

Object stiffness recognition and vibratory feedback without ad-hoc sensing on the Hannes prosthesis: A machine learning approach.

In recent years, hand prostheses achieved relevant improvements in term of both motor and functional recovery. However, the rate of devices abandonment, also due to their poor embodiment, is still high. The embodiment defines the integration of an external object - in this case a prosthetic device - into the body scheme of an individual. One of the limiting factors causing lack of embodiment is the absence of a direct interaction between user and environment. Many studies focused on the extraction of tactile information via custom electronic skin technologies coupled with dedicated haptic feedback, though increasing the complexity of the prosthetic system. Contrary wise, this paper stems from the authors' preliminary works on multi-body prosthetic hand modeling and the identification of possible intrinsic information to assess object stiffness during interaction. Based on these initial findings, this work presents the design, implementation and clinical validation of a novel real-time stiffness detection strategy, without ad-hoc sensing, based on a Non-linear Logistic Regression (NLR) classifier. This exploits the minimum grasp information available from an under-sensorized and under-actuated myoelectric prosthetic hand, Hannes. The NLR algorithm takes as input motor-side current, encoder position, and reference position of the hand and provides as output a classification of the grasped object (no-object, rigid object, and soft object). This information is then transmitted to the user via vibratory feedback to close the loop between user control and prosthesis interaction. This implementation was validated through a user study conducted both on able bodied subjects and amputees. The classifier achieved excellent performance in terms of F1Score (94.93%). Further, the able-bodied subjects and amputees were able to successfully detect the objects' stiffness with a F1Score of 94.08% and 86.41%, respectively, by using our proposed feedback strategy. This strategy allowed amputees to quickly recognize the objects' stiffness (response time of 2.82 s), indicating high intuitiveness, and it was overall appreciated as demonstrated by the questionnaire. Furthermore, an embodiment improvement was also obtained as highlighted by the proprioceptive drift toward the prosthesis (0.7 cm).

Fold2Vec: Towards a Statement-Based Representation of Code for Code Comprehension

We introduce a novel approach to source code representation to be used in combination with neural networks. Such a representation is designed to permit the production of a continuous vector for each code statement. In particular, we present how the representation is produced in the case of Java source code. We test our representation for three tasks:code summarization,statement separation, andcode search. We compare with the state-of-the-artnon-autoregressiveandend-to-endmodels for these tasks. We conclude that all tasks benefit from the proposed representation to boost their performance in terms of F1-score, accuracy, and mean reciprocal rank, respectively. Moreover, we show how models trained on code summarization and models trained on statement separation can be combined to address methods with tangled responsibilities, meaning that these models can be used to detect code misconduct.

Candidate project selection in cross project defect prediction using hybrid method

Cross Project Defect Prediction (CPDP) is a process that develops a defect prediction model on source projects and then applies the same model to the target project. Day by day, new software projects are being developed, so selecting an appropriate training project from existing projects and from new upcoming projects, to train a predictor model is a challenging task in CPDP. In the present study, we have proposed a hybrid selection method to select a candidate project from existing projects and a probabilistic method to select a candidate project from new projects. The proposed hybrid method is a weighted combination of the Collaborative filtering (CF) method and the Content Based (CB) method. The probabilistic method is based on a Naïve Bayes classifier and is used to predict the relation between the target project and the new target project. In the CF method, a usability score is generated for each project by making use of classification techniques, and the CB method calculates the matching score of candidate projects by using the K-dimensional tree. Finally, both the methods are combined by parallelized hybridization design, and weights for the proposed method are estimated with an empirical bootstrapping method. The score generated by the proposed hybrid technique is then used to identify the most suitable candidate project for the new project. The experimental results show that the suggestion list of the best three candidate projects is consistent when employing different classifiers. The recommendation performance is evaluated in terms of F-score and Mean Average Precision (MAP), and the proposed method has shown improved performance as compared to the existing methods in both terms.