Accurate traffic flow prediction remains a fundamental challenge in intelligent transportation systems, particularly in cross-domain, data-scarce scenarios where limited historical data hinders model training and generalisation. The complex spatio-temporal dependencies and nonlinear dynamics of urban mobility networks further complicate few-shot learning across different cities. This paper proposes MCPST, a novel Multi-phase Consensus Spatio-Temporal framework for few-shot traffic forecasting that reconceptualises traffic prediction as a multi-phase consensus learning problem. Our framework introduces three core innovations: (1) a multi-phase engine that models traffic dynamics through diffusion, synchronisation, and spectral embeddings for comprehensive dynamic characterisation; (2) an adaptive consensus mechanism that dynamically fuses phase-specific predictions while enforcing consistency; and (3) a structured meta-learning strategy for rapid adaptation to new cities with minimal data. We establish extensive theoretical guarantees, including representation theorems with bounded approximation errors and generalisation bounds for few-shot adaptation. Through experiments on four real-world datasets, MCPST achieves competitive performance against twenty-one (21) state-of-the-art methods in spatio-temporal graph learning, dynamic graph transfer learning, prompt-based spatio-temporal prediction, and cross-domain few-shot settings. Our framework delivers strong average prediction accuracy while requiring up to 90% less training data, and performs on par with the leading baselines in statistically significant comparisons. The implementation code is available at https://github.com/afofanah/MCPST.

Accurate traffic flow prediction is essential for intelligent transportation systems and urban planning. Traditional approaches that combine Transformer models with Graph Convolutional Networks (GCNs) or Convolutional Neural Networks (CNNs) often struggle to effectively integrate features with varying degrees of connectivity. As a result, graph-based problems do not fully utilise the capabilities of GCNs, while time-series problems fail to entirely leverage CNNs. To overcome these challenges, we introduce the Dual Heterogeneous Three-Stage Coupling and Multivariate Feature-Aware Learning Network (CHAMFormer). This architecture comprises three main components, each focusing on a distinct innovation: capturing fine-grained, short-range traffic patterns to manage immediate interactions and local bottlenecks; integrating mid-range spatial and temporal features to understand broader traffic interactions and ripple effects; and analysing complex, long-term traffic dynamics to anticipate and manage large-scale events and network behaviours across the entire system. These modules enhance GCN performance, enabling them to function more effectively alongside Transformers and Graph Neural Networks (GNNs). The CHAMFormer model incorporates a three-stage self-attention mechanism with a Skip-Connection method to improve the capture of detailed information without significantly increasing computational costs. By connecting low-level, intermediate-level, and high-level feature extractions, this model adapts well to changing traffic patterns, thereby enhancing multi-feature awareness and prediction accuracy. Extensive experiments using seven public datasets, both with and without predefined graph structures, and in multivariate and univariate scenarios demonstrate that CHAMFormer improves prediction accuracy by at least 10%–15%. To validate our proposed model, we also tested CHAMFormer in the energy domain, where it effectively handles time-series problems. Additionally, a sensitivity analysis confirms the model's predictability and interpretability, providing valuable insights for transportation policy and infrastructure development.

Significant challenges arise when Graph Neural Networks (GNNs) try to deal with uneven data. Specifically in signed and weighted graph structures. This makes classification tasks less effective. Within the GNN context, researchers have found traditional solutions like resampling, reweighting, and synthetic sample generation to be inadequate. GATE-GNN, a novel architecture designed specifically for imbalanced datasets, overcomes these limitations. GATE-GNN integrates an ensemble of network modules that harness the spatial features of graph networks and effectively utilise embedding information from earlier layers. This unique approach not only bolsters generalisation by reducing volatility. It also refines the optimisation algorithm, resulting in more accurate and stable classification outcomes. We rigorously tested the effectiveness of GATE-GNN on four widely recognised datasets: Cora, NELL, Citeseer, and PubMed. We performed a comparative analysis against established methods such as Graph Convolutional Networks (GCN), Graph Sample and Aggregate (GraphSAGE), Propagation Multilayer Perceptron PMLP), Imbalanced Node Sampling GNN (INS-GNN), GNN-Curriculum Learning (GNN-CL) and Graph Attention Networks (GAT). Empirical results demonstrate that GATE-GNN significantly outperforms these existing models, achieving an average improvement in classification accuracy of approximately 5%–10% over the previous best results. Additionally, GATE-GNN presents a marked reduction in training time. This underscores its efficiency and suitability for practical applications in imbalanced graph data scenarios. Implementation of the proposed GATE-GNN can be accessed here https://github.com/afofanah/GATE-GNN.

Abstract In this study, we evaluate the programming capabilities of OpenAI's GPT‐3.5 and GPT‐4 models using Swift‐based exam questions from a third‐year university course. The results indicate that both GPT models generally outperform the average student score, yet they do not consistently exceed the performance of the top students. This comparison highlights areas where the GPT models excel and where they fall short, providing a nuanced view of their current programming proficiency. The study also reveals surprising instances where GPT‐3.5 outperforms GPT‐4, suggesting complex variations in AI model capabilities. By providing a clear benchmark of GPT's programming skills in an academic context, our research contributes valuable insights for future advancements in AI programming education and underscores the need for continued development to fully realize AI's potential in educational settings.

This paper compares four prevalent neural network models (SNN, DNN, RNN, LSTM) in forecasting daily subway traffic volumes. We employ both real subway data and an artificially generated data set, the latter allowing for the calculation of a theoretical limit to forecast accuracy. The results indicate that LSTM outperforms the other models in accuracy across both data sets, while RNN displays a potential overfitting tendency. Our research contributes to the field by providing a benchmark for model accuracy and revealing unique insights about the performances of these models. The findings may aid in the efficient management of subway systems.

Abstract To improve their competitiveness, companies should acquire the latest software systems. However, many small or medium‐sized companies might not be able to afford a proper software acquisition process without the resource and technical support essential for acquiring the right software system. This article proposes a meta‐process, software acquisition with an external agent (SAEA), that includes a new industry role, external agent, to substitute for that support. SAEA will significantly reduce the cost of the software acquisition process. For implementing SAEA, the authors also introduce composition trees as a modelling language to collect, integrate, and compare all software requirements from different stakeholders and to identify issues during the software acquisition process. Finally, a real industry case study, for which this process has been successfully implemented, has been provided to illustrate our approach.

In the last two decades, several studies have been carried out to translate Behavior Trees (BTs) into other formal languages. However, as BTs are usually drawn directly from natural languages, there is no formal grammar to define what is a valid BT. In this research, we first propose a normal form for requirement BT as a building block for a valid BT, and then design a context-free grammar that can generate and verify all valid BTs. This work provides a solid foundation for BT research and will improve the quality of requirements modeling by identifying some common requirement defects.

Requirement Change is one of the most challenging tasks in software development lifecycle, particularly in the complex context of Global Software Development (GSD). During the last decade, many studies are carried out to address these problems, however, careful examination of these works suggests that there's a potential research gap. This paper has performed a Systematic Literature Review (SLR) to identify the most significant/commonly studied challenges of requirement change management process and furthermore this process under GSD context. We identified ten challenges such as impact analysis, cost estimation, artifacts documents management, requirement traceability, requirements dependency, conflicts with existing requirements, time estimation, change prioritization, user involvement, and system destabilizing. Furthermore, three challenges such as communication and coordination, knowledge sharing, management, and Change Control Board (CCB) management are identified for globally distributed projects. We also mapped these identified challenges to Requirement Change Management Process (RCMP) outcomes proposed in our previous study. We believe that mapping between RCM challenges and RCMP outcomes will enhance the practical significance of this study results. Considering the systematic literature review results, we suggest that there is a need to develop a framework for requirement change management for quality software systems development.

Spectral-spatial processing has been increasingly explored in remote sensing hyperspectral image classification. While extensive studies have focused on developing methods to improve the classification accuracy, experimental setting and design for method evaluation have drawn little attention. In the scope of supervised classification, we find that traditional experimental designs for spectral processing are often improperly used in the spectral-spatial processing context, leading to unfair or biased performance evaluation. This is especially the case when training and testing samples are randomly drawn from the same image - a practice that has been commonly adopted in the experiments. Under such setting, the dependence caused by overlap between the training and testing samples may be artificially enhanced by some spatial information processing methods, such as spatial filtering and morphological operation. Such enhancement of dependence in return amplifies the classification accuracy, leading to an improper evaluation of spectral-spatial classification techniques. Therefore, the widely adopted pixel-based random sampling strategy is not always suitable to evaluate spectral-spatial classification algorithms, because it is difficult to determine whether the improvement of classification accuracy is caused by incorporating spatial information into classifier or by increasing the overlap between training and testing samples. To tackle this problem, we propose a novel controlled random sampling strategy for spectral-spatial methods. It can greatly reduce the overlap between training and testing samples and provides more objective and accurate evaluation.

The critical behaviors of NP-complete problems have been studied extensively, and numerous results have been obtained for Boolean formula satisfiability (SAT) and constraint satisfaction (CSP), among others. However, few results are known for the critical behaviors of NP-hard nonmonotonic reasoning problems so far; in particular, a mathematical model for phase transition in nonmonotonic reasoning is still missing. In this article, we investigate the phase transition of negative two-literal logic programs under the answer-set semantics. We choose this class of logic programs since it is the simplest class for which the consistency problem of deciding if a program has an answer set is still NP-complete. We first introduce a new model, called quadratic model for generating random logic programs in this class. We then mathematically prove that the consistency problem for this class of logic programs exhibits a phase transition. Furthermore, the phase-transition follows an easy-hard-easy pattern. Given the correspondence between answer sets for negative two-literal programs and kernels for graphs, as a corollary, our result significantly generalizes de la Vega's well-known theorem for phase transition on the existence of kernels in random graphs. We also report some experimental results. Given our mathematical results, these experimental results are not really necessary. We include them here as they suggest that our phase-transition result is more general and likely holds for more general classes of logic programs.

In this paper, we present a semantic network model (SNM) as a reasoning engine for the requirements models. The SNM consists of the vertices and the edges, in which they store information of the models and their interrelations. The SNM, through a semi-automated normalisation process, helps the user (1) to assign states to the models and their relations as to whether they can be included, excluded, or undecided, (2) to eliminate redundant interrelations, (3) to avoid over-specification, and (4) to visualise a simplified overview of the whole system. Finally, we formulate the well-formedness of the SNM, which indicates whether the given models can produce a formal specification. We also evaluate our techniques using several case studies.

Abstract This paper proposes a model, the linear model, for randomly generating logic programs with low density of rules and investigates statistical properties of such random logic programs. It is mathematically shown that the average number of answer sets for a random program converges to a constant when the number of atoms approaches infinity. Several experimental results are also reported, which justify the suitability of the linear model. It is also experimentally shown that, under this model, the size distribution of answer sets for random programs tends to a normal distribution when the number of atoms is sufficiently large.

In this paper, we present a formal definition of the integration of the requirements modeling language Behavior Trees (BTs). We first provide the semantic integration of two interrelated BTs using an extended version of Communicating Sequential Processes. We then use a Semantic Network Model to capture a set of interrelated BTs, and develop algorithm to integrate them all into one BT. This formalisation facilitates developing (semi-)automated tools for modeling the requirements of large-scale software intensive systems.

Global Software Development projects could be best understood as intrinsically complex adaptive living systems: they can not purely be considered as 'designed systems', as deliberate design/ control episodes and processes (using 'software engineering' models) are intermixed with emergent change episodes and processes (that may perhaps be explained by models). Therefore the evolution of GSD projects includes the emergent as well as the deliberate aspects of system change. So to study GSD projects as complex systems we need to focus on both the state of the art of GSD research, as addressed in the software engineering discipline, as well as other disciplines that studied complexity such as Enterprise Architecture, Complexity and Information Theory, Axiomatic Design theory, for example. In this paper we study the complexity of GSD projects and propose the application of Extended Axiomatic Design (EAD) theory to reduce the complexity of GSD projects and to increase their probability of success. We also demonstrate that by satisfying all design axioms this 'structural' complexity could be minimised. By satisfying all three axioms of EAD, GSD management could make the life cycle activities of GSD planning and development projects as independent, controlled and uncoupled as possible so that the designer can predict the next relevant states of the life history and avoid a chaotic change in such projects.

In this paper we investigate how combining two types of modelling languages will increase their expressive power. The Behavior Tree method and non-monotonic logic will be integrated.

Complex-network theory is a new approach in studying different types of large systems in both the physical and the abstract worlds. In this paper, we have studied two kinds of network from software engineering: the component dependence network and the sorting comparison network (SCN). It is found that they both show the same scale-free property under certain conditions as complex networks in other fields. These results suggest that complex-network theory can be a useful approach to the study of software systems. The special properties of SCNs provide a more repeatable and deterministic way to study the evolution and optimization of complex networks. They also suggest that the closer a sorting algorithm is to the theoretical optimal limit, the more its SCN is like a scale-free network. This may also indicate that, to store and retrieve information efficiently, a concept network might need to be scale-free.

Complex-network theory is a new approach in studying different types of large systems in both the physical and the abstract worlds. In this paper, we have studied two kinds of network from software engineering: the component dependence network and the sorting comparison network (SCN). It is found that they both show the same scale-free property under certain conditions as complex networks in other fields. These results suggest that complex-network theory can be a useful approach to the study of software systems. The special properties of SCNs provide a more repeatable and deterministic way to study the evolution and optimization of complex networks. They also suggest that the closer a sorting algorithm is to the theoretical optimal limit, the more its SCN is like a scale-free network. This may also indicate that, to store and retrieve information efficiently, a concept network might need to be scale-free.

Estimating the parameters for a constant amplitude, polynomial phase signal with additive Gaussian noise is considered. The difficulty in this problem is that there are many unobserved integers when a linear regression model is used for wrapped phases. Analysing the least squares target function based on the regression model, we use the differencing approach to simplify it. Thus a tree-search algorithm can be used to find the solution of the least squares problem. To reduce the computational complexity, statistical inference methods are applied. Then an attractive recursive algorithm is derived. Simulation results show that this algorithm works at a lower SNR than that for existing methods.