In the field of software engineering, defect prediction has always been a popular research direction. Currently, the research on traditional software defect prediction mainly focuses on metric features, which are derived from various descriptive rules. Many researchers have proposed a large number of defect prediction models based on these metric features and various framework models. However, the problem of data scarcity has severely hindered the development of the field. Therefore, this work proposes a new method, namely the Metric Attention Module (MAM), which excavates the correlations within the metric data features, between features, within modules, and between modules. By learning new data representations, MAM guides the model's learning process and ultimately improves the model's performance without changing the network framework structure. Additionally, the method is interpretable. In this work, experiments were conducted in various task environments and on different datasets, all resulting in varying degrees of improvement. In the context of within-project defect prediction (WPDP), experiments with the MAM data model showed an average improvement of 14.7% in Accuracy, 15.9% in F1 score, 23.7% in AUC, and 65.1% in MCC. In cross-project defect prediction (CPDP), under more complex task environments, the model demonstrated excellent performance across multiple standard datasets. Compared to the baseline models and training results, the F1, Accuracy, and MCC scores improved by approximately 40%, 20%, and 50%, respectively.

Identification and definition of non-functional requirements (NFRs) are crucial for project success. Despite advancements have been achieved in automated requirements classification, it still suffers from data sparsity. To address these challenges, this paper proposes ChatNRC, a framework for NFR classification based on large language models (LLMs). The framework employs three LLMs to construct a generative-discriminative structure to mitigate data sparsity. Firstly, it utilizes the definitions of NFR categories and common feature words to guide Generative LLM to generate specific NFRs. To improve the diversity of generated requirements, specific prompts are utilized to guide the LLM in generating both generic and application-specific NFRs. Subsequently, Discriminative LLM discriminates the validity of generated requirements. Only those valid NFRs are included in the dataset to alleviate data sparsity and improve data diversity. Furthermore, the framework integrates the training dataset and the generated dataset to jointly train the NFR Classification LLM. Experimental results across the Security, Performance, Usability, and Operational NFR categories demonstrate that ChatNRC outperforms current state-of-the-art models on the PROMISE-NFR dataset, achieving a weighted average $\mathbf{F 1}$ score of $\mathbf{0 . 9 6}$. This demonstrates its superior performance and offers a novel approach for training LLMs in data-scarce domains.

GraphFusion is a hybrid question-answering framework that combines symbolic and semantic retrieval over a domain-specific knowledge graph (KG).Initially, schema-driven prompting guides a large language model (LLM) to generate precise Cypher queries for structured evidence extraction.If symbolic retrieval is insufficient, the system automatically employs semantic retrieval using transformer-based sentence embeddings.Both retrieval methods are integrated by the LLM within a Retrieval-Augmented Generation (RAG) pipeline, delivering coherent, factually accurate answers.Exhaustive schema inspection and parameterized queries ensure complete KG coverage, while FAISS indexing enables efficient semantic search.Evaluations on the RESQA and MetaQA benchmarks demonstrate that GraphFusion achieves strong in-domain accuracy and robust cross-domain generalization, highlighting the effectiveness of combining symbolic precision with neural semantic retrieval.

Knowledge Base Question Answering (KBQA) aims to find correct answers to natural language questions by reasoning over large-scale knowledge bases.The main challenge is multi-hop reasoning, which requires inferring answers through multiple edges and nodes.Due to high data annotation costs, most datasets only provide final answer annotations, leaving the reasoning process unknown.The model cannot provide the reasoning path along with the answer, making the answer uninterpretable.To address this, we propose a generation-retrieval multi-hop KBQA method combining large language models (LLMs) and the logic programming language Prolog.In the generation phase, we fine-tune the open-source LLMs to generate the logical form of the corresponding prolog representation for the natural language question.In the retrieval phase, the model uses prolog query to reasoning over the KB to get the final answer and the reasoning path.A transparent reasoning path also helps identify and correct errors, enhancing the model's reliability and practicality.Experimental results on two standard KBQA datasets, MetaQA and WebQSP, demonstrate that our method outperforms existing models in both performance and interpretability.

With the rapid development and widespread application of Knowledge graphs (KGs) in many artificial intelligence tasks, a large number of efforts have been made to refine them and increase their quality. Knowedge graph embedding (KGE) has become one of the main refinement tasks, which aims to predict missing facts based on existing ones in KGs. However, there are still mainly two difficult unresolved challenges: (i) how to leverage the local structural features of entities and the potential soft logical rules to learn more expressive embedding of entites and relations; and (ii) how to combine these two learning processes into one unified model. To conquer these problems, we propose a novel KGE model named JSSKGE, which can \textbf{J}ointly learn the local \textbf{S}tructural features of entities and \textbf{S}oft logical rules. Firstly, we employ graph attention networks which are specially designed for graph-structured data to aggregate the local structural information of nodes. Then, we utilize soft logical rules implicated in KGs as an expert to further rectify the embeddings of entities and relations. By jointly learning, we can obtain more informative embeddings to predict new facts. With experiments on four commonly used datasets, the JSSKGE obtains better performance than state-of-the-art approaches.

Recently, multivariate time series (MTS) clustering has gained lots of attention. However, state-of-the-art algorithms suffer from two major issues. First, few existing studies consider correlations and redundancies between variables of MTS data. Second, since different clusters usually exist in different intrinsic variables, how to efficiently enhance the performance by mining the intrinsic variables of a cluster is challenging work. To deal with these issues, we first propose a variable-weighted K-medoids clustering algorithm (VWKM) based on the importance of a variable for a cluster. In VWKM, the proposed variable weighting scheme could identify the important variables for a cluster, which can also provide knowledge and experience to related experts. Then, a Reverse nearest neighborhood-based density Peaks approach (RP) is proposed to handle the problem of initialization sensitivity of VWKM. Next, based on VWKM and the density peaks approach, an ensemble Clustering framework (SSEC) is advanced to further enhance the clustering performance. Experimental results on ten MTS datasets show that our method works well on MTS datasets and outperforms the state-of-the-art clustering ensemble approaches.

Traditional classification learning algorithms have several limitations: 1) they are time consuming for the large-scale training multivariate time-series (MTS) data, and unsuitable for the dynamically added training data; 2) as the number of the training MTS data becomes larger, they could not achieve the desired classification accuracy; 3) most of them do not consider how to make use of the unlabeled samples to enhance the classifier performance; and 4) due to the high dimension of MTS and complex relationship among variables, existing online learning algorithms are not effective to update shapelet-based association rules. Up to now, few work touched online classification learning for dynamically added unlabeled examples. To efficiently address these issues, we propose an online rule-based classifier learning framework on dynamically added unlabeled MTS data (ORCL-U). This framework integrates a confidence-based labeling strategy (CLS) and an online rule-based classifier learning approach (ORBCL). Extensive experiments on ten datasets show the effectiveness and efficiency of our proposed approach.

Recently, knowledge graph embedding approaches which aim to learn latent semantic representations for entities and relations in the knowledge graph, have become an active research topic. In this paper, we focus on hyperbolic knowledge graph embedding with logical pattern learning, based on the following observations: (i) the hyperbolic embedding methods have shown promising performance for knowledge graphs, which always exhibit hierarchical patterns; (ii) the Givens rotation has an excellent effect on modeling the logical pattern relations in the knowledge graph; (iii) ontology axioms are important extra information that can derive a variety of logical patterns (rules), and should be taken into consideration when learning embeddings. Futhermore, embedding learning and rule learning will complement each other and can be integrated into a unified framework. Therefore, we propose a novel hyperbolic knowledge graph embedding model named HyperLP, which can simultaneously learn embeddings and logical patterns. With extensive expriments on benchmark datasets, we find that our proposed HyperLP model achieves superior perfomence compared with a line of state-of-the-art baselines.

Semi-supervised clustering algorithms have several limitations: 1) the computation complexity of them is very high, because calculating the similarity distances of pairs of examples is time-consuming; 2) traditional semi-supervised clustering methods have not considered how to make full use of must-link and cannot-link constraints. In the clustering, the contribution of a few pairwise constraints to the clustering performance is very limited, and some may negatively affect the outcome; and 3) these methods are not effective to handle high dimensional data, especially for time series data. Up to now, few work touched semi-supervised clustering on time series data. To efficiently cluster large-scale time series data, we first tackle contract time series clustering to produce the most accurate clustering results under a contracted time. We propose a semi-supervised time series clustering framework (STSC), which integrates a fast similarity measure and a constraint propagation approach. Based on the proposed framework, two valid semi-supervised clustering algorithms including fssK-means and fssDBSCAN are designed. Experiments on 11 datasets show that our proposed method is efficient and effective for clustering large-scale time series data.

Protein-protein interaction (PPI) network alignment aims to find a node mapping between nodes across networks for different specifies such that the mapped nodes are topologically and biologically similar. Such alignment can help to reveal functionally similar subnetworks and predict previously undiscovered orthologs of different species. Although many aligners have been proposed, few can find global alignments with both high topological and biological quality. In this paper, we propose a new pairwise global PPI network aligner (GPPIAL) based on orthologs. We first construct a score matrix to evaluate the matching possibility of node pairs by combining multiple sources of information, including the biological (sequence and functional), topological (degree and neighborhood), and interaction information. Then we select orthologs with high similarity scores in the Inparanoid database as anchor pairs, and gradually build a global alignment. We compare GPPIAL with the stateof-the-art pairwise global network aligners on multiple PPI network pairs from the BioGRID dataset. The results show that GPPIAL significantly outperforms existing global aligners in terms of biological quality without losing the topological quality. Moreover, GPPIAL can also detect more orthologous proteins than existing global aligners, and thus can be applied to predicting the function of unannotated proteins. Our code and datasets are available at https://github.com/whuyrc/GPPIAL.

Trace links between requirements and software artifacts provide available traceability information and in-depth insights for different stakeholders. Unfortunately, establishing requirements trace links is a tedious, labor-intensive and fallible task. To alleviate this problem, Information Retrieval (IR) methods, such as Vector Space Model (VSM), Latent Semantic Indexing (LSI), and their variants, have been widely used to establish trace links automatically. But with the widespread use of agile development methodology, artifacts that can be used to generate automatic tracing links are getting shorter and shorter, which decreases the effects of traditional IR-based trace link generation methods. In this paper, Biterm Topic Model–Genetic Algorithm (BTM–GA), which is effective in managing short-text artifacts and configuring initial parameters, is introduced. A hybrid method VSM[Formula: see text]BTM–GA is proposed to generate requirements trace links. Empirical experiments conducted on five real and frequently-used datasets indicate that (1) the hybrid method VSM+BTM[Formula: see text]GA outperforms the others, and its results can achieve the “Good” level, where recall and precision are no less than 70% and 30%, respectively; (2) the performance of the hybrid method is stable and (3) BTM–GA can provide a number of “hard-to-find” trace links that complement the candidate trace links of VSM.

Embedding knowledge graphs (KGs) into continuous vector spaces is a focus of current research. Although previous works achieve great success, most of them are based on the closed-world assumption, where they only utilize knowledge explicitly exist in KGs but ignore implicit knowledge. This paper tries to improve the performance of KG embedding models by adding implicit knowledge to it. We consider that when an entity occurs only in head or tail of all triples in one KG, its connectivity with other entities is single and the inverse triples of it can provide as implicit knowledge to enrich its connectivity, and with this implicit knowledge KG embedding models can learn more accurate semantic of entities and relations in the KG. For this purpose, we introduce `fake triples' (triples that theoretically should exist but not appear in knowledge graph explicitly) via dummy relations for zero in-degree and zero out-degree entities to enrich their connectivity and further improve the embedding models' performance. Extensive experiments on entity alignment task and linking predication task show that our approach achieves good results. On entity alignment task, the entity alignment model with fake triples (EA+F) obtains better results than a number of state-of-the-art entity alignment models. On linking predication task, our method achieves better Mean Rank on FB15K. Our work can prove that the performance of knowledge graph embedding may be promoted by elaborate analysis on dataset rather than designing complex models.

Trace links between software artifacts provide available traceability information and in-depth insights for different stakeholders.Unfortunately, establishing trace links is a fallible, tedious, and labor-intensive task.To alleviate these problems, many Information Retrieval (IR) methods, such as Vector Space Model (VSM), Latent Semantic Indexing (LSI) and their variants, have been proposed to establish trace links automatically.In recent years, short-text artifacts (or even lack of documentation) become a new trend as more and more software systems are developed abiding by agile methodologies.It makes the effects of traditional IR-based trace links generation methods even worse.In this paper, Biterm Topic Model (BTM), which is good at dealing with short text, is introduced to solve the problem.A hybrid method combining VSM and BTM is proposed to generate requirements trace links.The empirical experiments conducted on three real and frequently-used datasets indicate that the hybrid method can achieve better performance, and the results can reach the "acceptable level" directly.

As cloud computing platforms are widely accepted by more and more enterprises and individuals, the underlying cloud workflow systems accumulate large numbers of business process models. Retrieving and recommending the most similar process models according to the tenant's requirements become extremely important, for it is not only beneficial to promote the reuse of the existing model assets, but also helpful to reduce the error rate of the modeling process. Since the scales of cloud workflow repositories become bigger and bigger, developing efficient and exact query approaches is urgent. To this end, an improved two-stage exact query approach based on graph structure is proposed. In the filtering stage, the composite task index, which consists of the label, join-attribute and split-attribute of a task, is adopted to acquire candidate models, which can greatly reduce the number of process models needed to be tested by a time-consuming verification algorithm. In the verification stage, a novel subgraph isomorphism test based on task code is proposed to refine the candidate model set. Experiments are conducted on six synthetic model sets and two real model sets. The results demonstrate that the presented approach can significantly improve the query efficiency and reduce the query response time.

In a business cloud environment, a reference business process model needs to be customized in order to meet the individualized requirements of each organization. Consequently, a reference process model is generally evolved to a couple of process variants, known as a process family. Currently, there exist some approaches and tools that can efficiently configure a reference process model. However, a key issue is how to manage co-evolution appropriately within a family of process models if the base process model of a process family is changed. Contemporary process management tools do not adequately support the management of such co-evolution. Each process variant in the process family has to be changed as a separate process model which often leads to redundancies and inefficiencies. In this article, we propose a novel approach for managing co-evolution of process families based on an aspect-oriented approach. Change options on base process model are abstracted and extracted as a pluggable component, which can be selectively reused for all members, and consequently, guides the co-evolution for the whole process family. In particular, the control-flow and data-flow relations between process extension and a member of process families can be built by leveraging process extensibility patterns. The correctness of our extension approach is proved based on graph theory. The whole approach in this article have been implemented as a working open-source prototype and tested against a real case study from the city logistic distribution domain as well as real data set from SAP reference models.

To achieve workflow retrieving and compliance checking based on behavior, it is necessary to identify the behavioral profiles of workflow models. Since the execution timing and the occurrence probability of tasks are crucial to choose appropriate workflows, it is significant to conduct exact quantitative analysis on the ordering relations by considering the time constraints and occurrence probabilities of tasks in workflow models. In this paper, the quantitative ordering relations are formally defined and an algorithm is proposed to efficiently compute the ordering relations with time and probability constraints based on complete firing sequences. Experiments show that the algorithm works well.

Requirements Elicitation (RE) is a critical process in system/software engineering. Its goal is to capture the stakeholders’ expectations, needs and constraints, which can be elicited, analyzed and specified as requirements. Gathering the requirements correctly, clearly and completely in a natural way is a typical challenging problem, because requirements analysts always play key roles in the elicitation process dominantly while stakeholders participate in passively. In this paper, we propose a collective intelligence driven business process oriented requirements acquisition and refining method. Its aim is to reduce the requirements analysts’ dominance and promote stakeholders’ self-expression and self-improvement to elicit requirements clearly and completely. It adopts the group storytelling method to promote the collaboration and communication among stakeholders, utilizes the narrative network model to enhance the associations among story fragments, and introduces dialogue game theory to guide the progressive refining. At the same time, the activity theory is adopted as the description framework to present the method and an application example is introduced. Finally, a pilot experiment is carried out to evaluate its perceived usefulness and perceived ease of use; and the actual quality of the requirements based on the degree of completeness and understandability in comparison with JAD. The results show that the requirements elicited by BPCRAR are more complete and understandable than JAD. In addition, the BPCRAR is perceived usefulness and ease of use in the experiment.

With the pervasiveness of the success cases of SaaS applications, outsourcing the manuscript submission and review system of periodical presses to SaaS providers becomes a new trend. This study proposes a solution for constructing the journal manuscript submission and review system based on level-3 maturity model of SaaS. In this article, the requirements of the system have been elicited, the architecture has been designed, and the key implementation has been presented. By renting the services provided by the system instead of buying or developing, the periodical presses can not only obtain better technical support, but also save costs. In addition, the periodical press can utilize its agglomeration effect based on SaaS platform and subscribe the value-added services such as journal recommendation service to implicit contributors and reviewer recommendation service to periodical presses which will be helpful to promote the quality and impact factor of the journal.

Wikis, as typical well-known knowledge management tools that support collaborative work, are adopted by more and more practitioners and researchers as the basis to develop Distributed Requirements Engineering (DRE) tools. Thus, many wikis which are enhanced specially for supporting various activities in Distributed Requirements Engineering (namely DRE-specific wikis) are developed. The main goal of this study is to discover all the available DRE-specific wikis, gain an insight into how and to what degree current DRE-specific wikis can support the DRE activities, and identify the future research directions. We adopt the methodology of systematic literature review to find DRE-specific wikis, analyze the features embodied in them, identify DRE activities supported by them, and cluster the users' feedbacks from their literatures. The results show that 1) distributed requirements elicitation is the most popular DRE activity supported by current DRE-specific wikis, 2) enhanced features are mainly designed for this phase, 3) the well recognized advantage for using DRE-specific wikis is that they can facilitate the collaborative work, and the disadvantages mainly lie in the organization of the content and the usability. Based on the findings of this review, the possible future research directions of DRE-specific wikis have been pointed out, especially in distributed requirements elicitation, negotiation, validation, and management. The importance of cross-over studies and empirical research are both emphasized at the end of the paper.

Currently there exist many kinds of semantic Web Service models on the internet. They are heterogeneous so that it is hard to understand and interoperate each other. In this paper, we study several mainstream semantic Web Service models, and extract meta-models for each from the perspective of semantic Web Service discovery. Based on this, we propose the universal meta-model for semantic Web Service registration within the background of meta-model framework for interoperability (MFI, ISO/IEC19763). Some cases are studied and infrastructure supporting our work is demonstrated. The work in the paper may be regarded as an extension to UDDI on service semantics facet, and has been proposed to ISO as ISO/IEC 19763 Part 7.

With the rapid development of service computing related technology, the number of application of service based system (SBS) in enterprise is increasing rapidly. Research on evolution of SBS is becoming more and more important. In this paper, we propose a taxonomy framework for evolution of SBS, which is illustrated from five perspectives: (a) motivations of SBS evolutionary changes (why), (b) stakeholders of SBS evolutionary changes (who), (c) locations of SBS evolutionary changes happening (where), (d) times of SBS evolutionary changes happening (when), (e) support mechanisms in the process of SBS evolutionary changes (how). Furthermore, propagation of evolutionary changes of SBS is analyzed in the paper.

In this paper, atomic services are orchestrated by a business process in the context of service composition. To enhance the quality of service composition, this paper introduces a behavior-based approach, which may alter the structure of business process aiming to preserving behavior semantics of the composite service. The result of the experiment shows that the quality of composite service can be improved in terms of performance time via the proposed approach. This paper presents a preliminary behavior-based restructure approach for service compositions to improve Quality of Service (QoS).

Software has become a complex piece of work by the collective efforts of many. And it is often hard to predict what the final outcome will be. This transition poses new challenge to the software engineering (SE) community. By employing methods from the study of complex network, we investigate the object oriented (OO) software metrics from a different perspective. We incorporate the weighted methods per class (WMC) metric into our definition of the weighted OO software coupling network as the node weight. Empirical results from four open source OO software demonstrate power law distribution of weight and a clear correlation between the weight and the out degree. According to its definition, it suggests uneven distribution of function among classes and a close correlation between the functionality of a class and the number of classes it depending on. Further experiment shows similar distribution also exists between average LCOM and WMC as well as out degree. These discoveries will help uncover the underlying mechanisms of software evolution and will be useful for SE to cope with the emerged complexity in software as well as efficient test cases design

Awareness of the physical location for each node is required by many wireless sensor network applications. The discovery of the position can be realized utilizing range measurements including received signal strength, time of arrival, time difference of arrival and angle of arrival. In this paper, we focus on localization techniques based on angle of arrival information between neighbor nodes. We propose a new localization and orientation scheme that considers beacon information multiple hops away. The scheme is derived under the assumption of noisy angle measurements. We show that the proposed method achieves very good accuracy and precision despite inaccurate angle measurements and a small number of beacons

Localization is a fundamental service for many applications in wireless sensor networks. In this paper we propose a probabilistic, constraint-based approach robust to range measurement inaccuracies. The proposed approach pro- ceeds in three phases: the first phase involves modeling the uncertainties of range measurements; in the second phase, a set of probabilistic constraints are computed and combined to produce initial position estimates; in the final phase, negative constraints are used to refine the initial estimates. We evaluated the proposed approach through simulations based on real-world measurements; the results are compared with two other local- ization schemes and the Cramer-Rao lower bound. The results show that, for inaccurate range measurements, the proposed probabilistic approach performs the best and close to the optimal bound.

Lattice low-delay vector excitation coding (LLD-VXC) is a speech coding system based on analysis-bysynthesis excitation coding and backward adaptation of the synthesis filter. The introduction of a lattice filter as a (high order) short-term predictor has significant advantages, such as fast tracking of speech signal nonstationarities, simple stability verification, and uniform distribution of the computational load. The objective of this paper is to present a Lattice LD-VXC (LLD-VXC) codec and experimental results obtained at rates of 8, 9.6, and 16 kb/s. A sign algorithm for lattice filter adaptation is introduced in order to reduce computational complexity. An LLD-VXC codec with a 20th-order lattice predictor, a 10thorder lattice weighting filter, and a backward pitch predictor achieved toll quality at 16 kb/s and good communications quality at 8-9.6 kb/s with a delay of less than 2 ms and reasonable complexity.

The authors report results on lattice low delay vector excitation coding (LLD-VXC) for rates in the range of 8-16 kb/s. The LLD-VXC, which is based on a backward adaptive analysis-by-synthesis configuration, an adaptive lattice short-term predictor, an adaptive lattice perceptual weighting filter, and an adaptive pitch predictor, offers toll speech quality at 16 kb/s with moderate complexity and a total communication delay of under 2 ms. For bit rates below 16 kb/s, the LLD-VXC keeps the same basic structures and algorithms but uses different code-vector dimensions and codebook sizes to achieve different bit rates. For reducing computational complexity, the sign-error-sign-data least mean square adaptive lattice algorithm was derived and implemented in LLD-VXC. The sign algorithm significantly reduced computational complexity at the expense of a slight speech quality degradation.>

The convergence properties of an adaptive lattice filter using a stochastic gradient algorithm are investigated using differential equations. The mean of the PARCOR coefficients of the adaptive lattice filter is obtained by analyzing an associated ordinary differential equation (ODE). An efficient way to compute the statistics required for the solution of the ODE is presented. An expression for the variance of the PARCOR coefficients is derived from the stochastic differential equation (SDE) associated with the normalized error process. Simulation results are given to support the theoretical results.>

In this paper we study the problem of adaptively estimating the Autoregressive Moving Average (ARMA) parameters of a time varying ARMA process using a constant step size Gauss-Newton Algorithm. Using weak convergence theory and the concept of prescaling, it is shown that the "mean" behavior can be described by an ordinary differential equation (ODE). Computer simulations are provided to substantiate the analysis.