Intelligent Information Processing XII

Anomaly detection in attributed networks has been crucial in many critical domains and has gained significant attention in recent years. However, most existing methods fail to capture the complexity of anomalous patterns at different levels with suitable supervision signals. To address this issue, we propose a novel dual contrastive self-supervised learning method for attributed network anomaly detection. Specifically, our approach relies on two major components to determine the anomaly of nodes. The first component assesses self-consistency by determining whether a target node’s attributes are consistent with its contextual environment. The second component evaluates behavioral consistency by analyzing the relationships and interaction patterns between the target node and its one-hop neighbors, which determines if the behavior of these neighbors aligns with the expected pattern of the target node. Accordingly, our method designs two types of contrastive instance pairs to fully exploit the structural and attribute information for detecting anomalous nodes at different levels regarding two focused consistencies. This approach is more effective in detecting anomalies and mitigating the limitations of previous methods. We evaluated our method on six benchmark datasets, and the experimental results demonstrate the superiority of our methods against state-of-the-art methods.

Most conventional literature on online learning implicitly assumes a static feature space. However, in real-world applications, the feature space may vary over time due to the emergence of new features and the vanishing of outdated features. This phenomenon is referred to as online learning with Varying Feature Space (VFS). Recently, there has been increasing attention towards exploring this online learning paradigm. However, none of the existing approaches have taken into account the potentially informative information conveyed by the presence or absence (i.e., variation in this paper) of each feature. This indicates that the existence of certain features in the VFS can be correlated with the class labels. If properly utilized for the learning process, such information can potentially enhance predictive performance. To this end, we formally define and present a learning framework to address this specific learning scenario, which we refer to as Online learning in Varying Feature space with Informative Variation (abbreviated as OVFIV). The framework aims to answer two key questions: how to learn a model that captures the association between the existence of features and the class labels, and how to incorporate this information into the prediction process to improve performance. The validity of our proposed method is verified through theoretical analyses and empirical studies conducted on 17 datasets from diverse fields.

With the rapid development of Deep Learning, more and more applications on the cloud and edge tend to utilize large DNN (Deep Neural Network) models for improved task execution efficiency as well as decision-making quality. Due to memory constraints, models are commonly optimized using compression, pruning, and partitioning algorithms to become deployable onto resource-constrained devices. As the conditions in the computational platform change dynamically, the deployed optimization algorithms should accordingly adapt their solutions. To perform frequent evaluations of these solutions in a timely fashion, RMs (Regression Models) are commonly trained to predict the relevant solution quality metrics, such as the resulted DNN module inference latency, which is the focus of this paper. Existing prediction frameworks specify different RM training workflows, but none of them allow flexible configurations of the input parameters (e.g., batch size, device utilization rate) and of the selected RMs for different modules. In this paper, a deep learning module inference latency prediction framework is proposed, which i) hosts a set of customizable input parameters to train multiple different RMs per DNN module (e.g., convolutional layer) with self-generated datasets, and ii) automatically selects a set of trained RMs leading to the highest possible overall prediction accuracy, while keeping the prediction time/space consumption as low as possible. Furthermore, a new RM, namely MEDN (Multi-task Encoder-Decoder Network), is proposed as an alternative solution. Comprehensive experiment results show that MEDN is fast and lightweight, and capable of achieving the highest overall prediction accuracy and R-squared value. The Time/Space-efficient Auto-selection algorithm also manages to improve the overall accuracy by 2.5% and R-squared by 0.39%, compared to the MEDN single-selection scheme.

Tables are commonly used for structuring and consolidating knowledge, significantly enhancing the efficiency for human readers to acquire relevant information. However, due to their diverse structures and open domains, employing computational methods for their automatic analysis remains a substantial challenge. Among these challenges, accurately classifying the forms of tables is fundamental for achieving deep comprehension and analysis, forming the basis for understanding, retrieving, and extracting knowledge within tables. Common table formats include row tables, column tables, and matrix tables, where data is arranged in rows, columns, and combinations of rows and columns, respectively. This paper introduces a novel approach for table classification based on the neural network model, TableTC. TableTC initially utilizes fine-tuning of the BERT pre-trained model to comprehend table content. Additionally, it proposes an improved Temporal Convolutional Network (TCN) named Temporal Convolutional Sparse Multilayer Perceptron Network (TCSMN). This network captures sequential structural features of cells and their surrounding neighbors, enhancing the ability to extract semantic features and positions. Finally, it employs an attention mechanism to further augment the capability of extracting row-column positions and semantic features. The evaluation of our proposed method is conducted using table data from scientific literature found in the PubMed Central website. Experimental results demonstrate that TableTC achieves a 2.7% improvement in table classification accuracy, as measured by the F1 score, compared to previous state-of-the-art methods on this dataset.

Dynamic Bayesian Networks (DBNs), renowned for their interpretability, have become increasingly vital in representing complex stochastic processes in various domains such as gene expression analysis, healthcare, and traffic prediction. Structure learning of DBNs from data is a challenging endeavor, particularly for datasets with thousands of variables. Most current algorithms for DBN structure learning are adaptations from those used in static Bayesian Networks (BNs), and are typically focused on smaller-scale problems. In order to solve large-scale problems while taking full advantage of existing algorithms, this paper introduces a novel divide-and-conquer strategy, originally developed for static BNs, and adapts it for large-scale DBN structure learning. Additionally, we leverage the prior knowledge of 2 Time-sliced BNs (2-TBNs), a special class of DBNs, to enhance the performance of this strategy. Our approach significantly improves the scalability and accuracy of 2-TBN structure learning. Designed experiments demonstrate the effectiveness of our method, showing substantial improvements over existing algorithms in both computational efficiency and structure learning accuracy. In problem instances with more than 1,000 variables, our proposed approach on average improves two accuracy metrics by $$74.45\%$$ 74.45 % and $$110.94\%$$ 110.94 % , respectively, while reducing runtime by an average of $$93.65\%$$ 93.65 % . Moreover, in problem instances with more than 10,000 variables, our proposed approach successfully completed the task in a matter of hours, whereas the baseline algorithm failed to produce a reasonable result within a one-day runtime limit.

Explainable reinforcement learning has evolved rapidly over the years because transparency of the model’s decision-making process is crucial in some important domains. Differentiable decision trees have been applied to this field due to their performance and interpretability. However, the number of parameters per branch node of a differentiable decision tree is related to the state dimension. When the feature dimension of states increases, the number of states considered by the model in each branch node decision also increases linearly, which increases the difficulty of human understanding. This paper proposes a entroy-based differentiable decision tree, which can restrict each branch node to use as few features as possible to predict during the training process. After the training is completed, the parameters that have little impact on the output of the branch node will be blocked, thus significantly reducing the decision complexity of each branch node. Experiments in multiple environments demonstrate the significant interpretability advantage of our proposed approach.

Deep clustering has powerful capabilities of dimensionality reduction and non-linear feature extraction, superior to conventional shallow clustering algorithms. Deep learning and clustering can be unified through one objective function, significantly improving clustering performance. However, the features of embedding space may have redundancy and ignore preserved manifold. Besides, the features lack discriminative, which hinders the clustering performance. To solve the above problems, the paper proposes a novel algorithm that improves the discrimination of features, filters redundant features and protects manifold structures for clustering. Firstly, it reduces the dimensionality in the embedding again to filter redundant and preserve the manifold for the features. Then it improves the discriminative of the representation by reducing the intra-class distance. Performance evaluation is carried out on four benchmark datasets and a case study of engineering applications. Comparing with state-of-the-art algorithms indicates that our algorithm performs favorably and demonstrates good potential for real-world applications.

Aiming at the problems of strict preference judgment and cold start in Bayesian personalized ranking(BPR), an improved ranking model is proposed, which considers the influence of time and incorporates hot recommendations. By extracting user behavior features, constructing an optimized BPR model, and processing recommendation results, we establish BPR-TH for realizing personalized online (or offline) recommendation of digital library information. By Comparing with other two similar algorithms, the experimental results show that this model performs better.

This paper aims to explore the concept of “depth” through the selection of various ensemble methods and proposes a practical deep ensemble learning method. In this study, we propose a nested ensemble learning method. First, we employ the stacking framework for selective ensemble learning. Next, we integrate the stacked ensemble with bagging and boosting techniques to create a comprehensive stacked ensemble. We utilized both domestic and foreign online loan data to build the model and test its ability to generalize. The experimental results demonstrate that the nested ensemble proposed in this paper outperforms models such as logistic regression and support vector machines, showing exceptional generalization ability.

Generative Adversarial Networks (GANs) have obtained outstanding performance in image-to-image translation. Nevertheless, their applications are greatly limited due to high computational costs. Although past work on compressed GANs has yielded rich results, most still come at the expense of image quality. Therefore, in order to generate high-quality images and simplify the process of distillation, we propose a framework with more generators and fewer discriminators (MGFD) strategy to enhance the online knowledge distillation with high-quality images. First, we introduce the Inception-enhanced residual block into our enhanced teacher generator, which significantly improves image quality at a low cost. Then, the multi-granularity online knowledge distillation method is adopted and simplified by selecting wider Inception-enhanced teacher generator. In addition, we also combine the intermediate layer distillation losses to help student generator to obtain diverse features and more supervised signals from the intermediate layer for better transformations. Experiments demonstrate that our framework can significantly reduce computational costs and generate more natural images.

Based on the theories of Riemannian surface, Topology and Analytic function, a novel method for dimensionality reduction is proposed in this paper. This approach utilizes FCA to merge highly correlated features to obtain approximate independent new features in the locally, and establishes a conformal homomorphic function to realize global dimensionality reduction for text data with the manifold embed in the Hausdorff space. During the process of dimensionality reduction, the geometric topological structure information of the original data is preserved through conformal homomorphism function. This method is characterized by its simplicity, effectiveness, low complexity, and it avoids the neighbor problem in nonlinear dimensionality reduction and it is conducive to the outlier data. Moreover, it has extensible for new text vectors and new feature from sub-vectors of new text vectors, and incremental operation without involving existing documents. The mapping function exhibits desirable properties resulting in stable, reliable, and interpretable dimensionality reduction outcomes. Experimental results on both construction laws and regulations dataset and toutiao text dataset demonstrate that this dimensionality reduction technique is effective when combined with the typical classification method of Random Forest, Support Vector Machine, and Feedforward Neural Network.

It is commonly seen that the imperfect multi-modal data with missing modality appears in realistic application scenarios, which usually break the data completeness assumption of multi-modal analysis. Therefore, large efforts in multi-modal learning communities have been made on the robust solution for modality-missing data. Recently, pre-trained models based on Mixture-of-Modality-Experts (MoME) Transformers have been proposed, which achieved competitive performance in various downstream tasks, by utilizing different experts of feed-forward networks for single/multi modal inputs. One natural question arises: are Mixture-of-Modality-Experts Transformers robust to missing modality? To that end, in this paper, we conduct a deep investigation on MoME Transformer under the missing modality problem. Specifically, we propose a novel multi-task learning strategy, which leverages a uniform model to handle missing modalities during training and inference. In this way, the MoME Transformer will be empowered with robustness to missing modality. To validate the effectiveness of our proposed method, we conduct extensive experiments on three popular datasets, which indicate our method could outperform the state-of-the-art (SOTA) methods with a large margin.

Although Pre-trained Language Model (PLM) ChatGPT as a Question-Answering System (QAS) is so successful, it is still necessary to study further the QASs based on PLMs. In this paper, we survey state-of-the-art systems of this kind, identify the issues that current researchers are concerned about, explore various PLM-based methods for addressing them, and compare their pros and cons. We also discuss the datasets used for fine-tuning the corresponding PLMs and evaluating these PLM-based methods. Moreover, we summarise the criteria for evaluating these methods and compare their performance against these criteria. Finally, based on our analysis of the state-of-the-art PLM-based methods for QA, we identify some challenges for future research.

Legal documents require high precision and accuracy in language use, leaving no room for grammatical and spelling errors. To address the issue, this paper proposes a novel application of the BERT pre-trained language model for legal document proofreading. The BERT-based model is trained to detect and correct legal texts’ grammatical and spelling errors. On a dataset of annotated legal documents, we experimentally show that our BERT-based model significantly outperforms state-of-the-art proofreading models in precision, recall, and F1 score, showing its potential as a valuable tool in legal document preparation and revision processes. The application of such advanced deep learning techniques could revolutionise the field of legal document proofreading, enhancing accuracy and efficiency.

Entity relation extraction aims to identify entities and their semantic relationships from unstructured text. To address issues like cascading errors and redundant information found in current joint extraction methods, a One-Module One-Step model is adopted. Additionally, in overcoming challenges related to limited annotated data and the tendency of neural networks to overfit, this paper introduces a method leveraging data augmentation based on a large language model. The approach utilizes five data augmentation strategies to improve the accuracy of triple extraction. Conducting experiments on the augmented dataset reveals significant enhancements in evaluation metrics compared to unaugmented data. In entity relation extraction tasks, the proposed method demonstrates a notable boost, increasing accuracy and F1 scores by 7.3 and 8.5 percentage points, respectively. Moreover, it shows a positive impact on the non-prompting strategy, elevating accuracy and F1 scores by 9.4 and 9.1 percentage points, respectively. These experiments affirm the effectiveness of data augmentation based on a large language model in improving entity relation extraction tasks.

It has been proven that non-elitist evolutionary algorithms (EAs) with proper selection mechanisms, including the recently proposed power-law ranking selection, can efficiently escape local optima on a broad class of problems called SparseLocalOpt $$_{\alpha ,\varepsilon }$$ α , ε , where elitist EAs fail. However, those theoretical upper bounds on the runtime are not tight as they require large populations and a tight balance between mutation rates and selection pressure to keep the algorithms operating near the so-called “error threshold”. This paper empirically clarifies the significance of these theoretical requirements and makes a series of performance comparisons between the non-elitist EA using power-law ranking selection and other EAs on various benchmark problems.Our experimental results show that non-elitist EAs optimise the Funnel problem with deceptive local optimum significantly faster with power-law ranking selection than with tournament selection. Furthermore, power-law selection outperforms UMDA and the (1+1) EA in our experiments on the NK-Landscape and Max k-Sat problems, but yields to the $$(\mu ,\lambda )$$ ( μ , λ ) -selection, tournament selection, and the self-adaptive MOSA-EA. On the unicost set cover problems, the EA with power-law selection shows competitive results.

As a special case of the multiobjective optimization problem, the multiobjective knapsack problem (MOKP) widely exists in real-world applications. Currently, most algorithms used to solve MOKPs assume that these problems involve only one decision maker (DM). However, some complex MOKPs often involve more than one decision makers and we call such problems multiparty multiobjective knapsack problems (MPMOKPs). Existing algorithms cannot solve MPMOKPs effectively. To the best of our knowledge, there is only a little attention paid to MPMOKPs. In this paper, inspired by existing SMS-EMOA, we propose a novel indicator-based algorithm called SMS-MPEMOA to solve MPMOKPs, which aims to search solutions to satisfy all decision makers as much as possible. SMS-MPEMOA is compared with several state-of-the-art multiparty multiobjective optimization algorithms (MPMOEAs) on the benchmarks and the experimental results demonstrate that SMS-MPEMOA is very competitive.

Many-objective Optimization problems (MaOPs), with four or more objectives are difficult to solve, is a kind of common optimization problems in actual industrial production. In recent years, a large number of many-objective optimization algorithms (MaOEAs) have been proposed to solve various types of MaOPs. However, in practical problems, it is usually hard to improve the existing optimization algorithms or make a lot of attempts for MaOEAs because the true Pareto surface is usually unknown in a new MaOPs, which is a time-consuming and uncertain task. In this paper, inspired by the selective hyper heuristic optimization algorithm, we propose an integrated hyper-heuristic many-objective optimization algorithm (MaOEA-EH), which can integrate the existing advanced MaOEAs by simulating the PBFT consensus mechanism in the blockchain, and select the best algorithm for the current problem through the voting-election method in the iterative process. Numerical results show that our algorithm performs well on various many-objective problems.

The Snake game, a classic in the gaming world, gains new dimensions with the Snake AI competition, where two players controlled by AI algorithms can now compete simultaneously in the same game session. This competition holds significance in advancing our understanding of artificial intelligence (AI) algorithms. In the 2020 and 2021 Snake AI competitions, popular algorithms, using graph-based search or heuristic strategies, demonstrate competitive performance, such as the A* algorithm, Monte Carlo Tree Search (MCTS). Contrary to these heuristic approaches, the Rolling Horizon Co-evolution Algorithm (RHCA), characterised by its core principles of rolling horizon evaluation and co-evolution, maintains two populations, one for each player, to co-evolve with each other without reliance on heuristics. RHCA has been verified its effectiveness in a two-player spaceship game. In this paper, we extend the RHCA application to the two-player Snake AI game, comparing it with other state-of-the-art methods. Additionally, we introduce various obstacles to create different complex scenarios, ensuring a comprehensive analysis. Experimental results reveal RHCA’s superior and stable performance, especially in resource-constrained and complex scenarios. Furthermore, an analysis of RHCA’s behaviours across maps with diverse obstacle scenarios highlights its ability to make intelligent decisions in competing with state-of-the-art methods.

In recent years, generative modelling has become a significant area of computer science research and artificial intelligence. This has been primarily due to the fact that generative models are useful in addressing the class imbalance problem inherent in some datasets. By generating synthetic data samples for underrepresented classes with a decent amount of variation through random noise, classification models could be trained more efficiently. The popularity of generative models was also increased by the prospect of being able to generate previously non-existent samples of images, audio and video for other creative tasks not related to addressing the class imbalance in datasets. This paper presents exploratory research to train an artificial immune network as a standalone generative model (called a generative adversarial artificial immune network, or GAAINet) using purely immunological computation concepts, such as antibody affinity, clonal selection and hypermutation. Experimental results show that the resulting generator artificial immune network could generate human-recognisable synthetic handwritten digits without any prior knowledge of the MNIST handwritten digits dataset.

Wide-channel plate heat exchanger is a widely-used high performance heat exchanger, and its structure has a significant effect on heat exchange effect. However, the density and flow rate of the heat transfer medium is uncertain, and we only can obtain their possible ranges. Based on this, interval number is introduced to describe uncertainty factor, and then formulate the interval constraint of wide-channel plate heat exchanger. The triangular fuzzy number is employed to define the degree of constraint violation. Due to the difficulty of modeling heat exchange efficiency, its surrogate model is trained by neural network. To solve this issue, multi-objective particle swarm optimization algorithm is developed to find the optimal structural variable of heat exchanger under uncertain conditions. The experimental results indicate that the proposed algorithm obtains the structure variable of heat exchanger with the most preferable heat effect and lowest cost quickly.

The scheduling optimization of industrial processes is crucial for enhancing production capacity and minimizing energy consumption. In the realm of continuous casting, the expansion of the scheduling scale and the increasing number of scheduling objects pose challenges for genetic algorithms in swiftly generating optimal solutions that adhere to constraints. Prolonged scheduling decision times and difficulties in ensuring constant pouring constraints are critical issues that require urgent resolution in the continuous casting scheduling problem within steelmaking. This paper proposes a genetic algorithm driven by translational mutation operator for the scheduling optimization in the steelmaking-continuous casting production named TMGA. Incorporating continuous pouring information in the encoding process guarantees uninterrupted pouring during the casting stage. Furthermore, applying the translational mutation operator is instrumental in elevating the search efficiency for the global optimal solution, consequently diminishing scheduling decision times. To validate the effectiveness of the proposed approach, this study conducts a rigorous examination involving a numerical simulation case and two ablation experiments. The experimental results demonstrate the superior performance of TMGA compared to other methods.

Modern computing and networking environments provide the important problems of efficient using such resources as energy and cores or processors. It is based on the possibility of dynamically varying the speed of processors and using parallel calculations in the execution of operations. We consider the NP-hard speed scaling scheduling problem with energy constraints and parallelizable jobs. Each job must be executed on the given number of processors. Processors can vary their speeds dynamically. It is required to assign speeds to jobs and schedule them such that the total completion time is minimized under the given energy budget. An adaptive genetic algorithm with optimized crossover operators is proposed. The optimal recombination problem is solved in the crossover operator. This problem is aimed at searching for the best possible offspring following the well-known gene transmitting property. The experimental evaluation shows that the algorithm outperforms the known metaheuristics and demonstrates the perspectives of using adaptive techniques and optimized operators.

Whole-brain network modeling (WBM) offers a pivotal tool to explore the large-scale spatiotemporal dynamics of the brain at rest, during cognitive tasks, and under external stimulation. However, it is unclear how to fuse multi-modal neural dynamics in a united WBM framework and predict the whole-brain spatiotemporal neural responses to electrical stimulation. In this study, we present a computational framework with whole-brain network modeling, parameter optimization, and model validation using simultaneous EEG-SEEG data during intracranial brain stimulation. To test the efficacy of WBM in revealing brain-wide neural dynamics, our experiments utilize synthetic electrophysiological data, real EEG data, and real EEG-SEEG signals. Experimental results demonstrate that our WBM framework accurately captures the spatiotemporal brain activities by jointly leveraging the higher spatial resolution from SEEG and the whole-brain coverage from EEG. Notably, our model shows a higher correlation between the functional connectivity (FC) matrix of EEG and that of the inferred whole-brain neural dynamics from WBM (r=0.86), compared to the FC from EEG source localization (r=0.48). Together, we demonstrate the capability and flexibility of WBM framework to uncover the whole-brain spatiotemporal neural activity and its potential to provide new insights into the input-response mechanism of the brain.

The exchange of real-time data between vehicle-to-vehicle communications is crucial in the Internet of Vehicles (IoV) for vehicle-intelligent decisions. However, malicious and false communication data may cause serious personal safety accidents. Confirming the authenticity of the identities of both parties and encrypting communication content before communication is the first line of defense to ensure system security. Therefore, to secure the vehicle-to-vehicle communications, this paper proposes a secure and efficient authentication and key agreement scheme with lightweight operation. Our scheme achieves vehicle-to-vehicle authentication and establishes a session key to encrypt subsequent communication content with only lightweight operations such as symmetric encryption algorithms and hash functions. Furthermore, our scheme provides many ideal attributes, such as forward secrecy, which ensures that the final compromised of the system will not affect the previous communication content. Besides, we prove the security of the proposed scheme through heuristic analysis and BAN logic analysis and analyze the performance of the proposed scheme via comparing the computational cost and communication cost with three state-of-the-art related schemes. The results show that the proposed scheme has high communication efficiency.

Social networks not only help expand interpersonal interactions, enable data analysis, and implement intelligent recommendations, but also can deeply examine social structures and dynamic changes between individuals, making them an indispensable part of contemporary society. However, malicious entities pose a significant threat to user identity and relationship information within social networks, raising concerns about privacy and security issues. Although existing k-anonymity schemes provide certain privacy protection, they lack the flexibility to adjust the intensity of privacy protection according to specific scenarios and user preferences, thus seriously compromising the utility of anonymized data. Based on the isomorphic algorithm, this paper proposes a new structural anonymity algorithm called α-partial isomorphic anonymity (α-PIA) to meet the privacy protection and data usage requirements in different scenarios of social networks. By capturing graph structure features at different levels to calculate the similarity between nodes, α-PIA can improve clustering quality. Extensive experiments are carried out based on two public datasets. Experimental results show that compared with similar schemes, α-PIA achieves better results in terms of information loss, average clustering coefficient and average shortest path length and better balances the privacy protection and practicality of graph data.

Knowledge Graph (KG) contains rich semantic information and supports knowledge reasoning. In recent years, introducing KG as auxiliary information into the recommender system has become one common measure for improving recommendation quality. The unified graph, which is constructed from the KG and user-item matrix in recommender systems, contains meta-paths formed by single-hop/continuous multi-hop connectivity relationships, and these meta-paths can assist modeling of user preferences. The quality of manually designed meta-paths is prone to the type and number of human-defined meta-paths. Moreover, the process of defining meta-paths is time-consuming and labor-intensive, and inadequate sufficient considerations in design will have an adverse impact on the quality of recommendations. We propose a self-supervised meta-path generation approach that does not rely on domain knowledge to select valuable path information from the unified graph and can deliver high-quality recommendations and reduce noises. Previous studies on meta-paths mainly focused on the neighbor information of nodes and ignored the edges that represents relationships between nodes. We develop a meta-path-based relational path-aware strategy to discover the relational information included within the meta-path. To make the use of the global structure in the unified graph and the information within the local scope in the user-item bipartite graph and KG, a two-level relationship aggregator to fully aggregate the fine-grained semantic information and multi-hop semantic associations is also proposed. We conducted experiments on two public datasets, MovieLens and Book-Crossing to verify the effectiveness of the proposed algorithm. The experimental results show that the recommendation algorithm outperforms the baseline models in terms of AUC, Recall@K, and F1 in most cases.

The itinerary planning problem plays a pivotal role in the tourism industry, involving the selection of an optimal tour route from multiple preferred points of interest (POIs) chosen by travelers while considering their diverse needs. However, as tourism expands and transportation becomes more accessible, there is a growing preference among travelers for planning single trips across multiple cities-referred to as cross-city itinerary planning. This paper introduces a novel approach, called CCIP, the cooperative coevolution framework for cross-city itinerary planning, which employs a divide-and-conquer method to automatically devise scalable cross-city itineraries, accounting for travelers’ preferences regarding time and travel choices. Experimental evaluations on real datasets from various cities in Jiangsu Province demonstrate that the proposed algorithm outperforms two classical multi-objective optimization algorithms, as measured by the HV metric.

Investor sentiment, as a factor influencing stock price volatility, has received increasing research attention in recent years. This study proposes a more comprehensive representation of sentiment by incorporating social attributes when constructing investor factors. Notably, a novel market sentiment factor, γ, is introduced in this paper, which combines investor sentiment, stock data, and policy influences to enhance prediction accuracy beyond individual models. A multi-level nested ensemble model based on stacking is constructed in this study, which integrates the sentiment-stock Dual-CNN-LSTM model with learners to improve the accuracy of stock price volatility prediction. The experimental results demonstrate that: (1) The proposed market sentiment factor γ shows improved predictive accuracy compared to using investor sentiment factors alone, with an average increase of 5.55%; (2) The Dual-CNN-LSTM model outperforms the CNN-LSTM model using stock data alone in terms of volatility prediction accuracy, with an improvement of 9.81%. (3) The proposed multi-level nested ensemble algorithm, which adopts stacking nested Learner, achieves an accuracy of 88.24% in stock trend prediction. Overall, this research constructs a better sentiment indicator factor γ and provides a new approach for predicting stock price volatility through the integrated nested model, highlighting the effectiveness of hybrid architectures in addressing financial forecasting challenges.

P2P, as the most representative online lending platform with a long history of personal credit development, can provide powerful data support for exploring the problem of personal credit default risk, and Logistic Regression plays an important role in machine learning, and the current research on Logistic Regression mainly stays at the application level. Therefore, based on the Factor Space theory to further deepen the interpretation of Logistic Regression, explore the obvious and hidden relationship of the factors behind it, and give a reasonable expression of Logistic Regression from the perspective of the obvious and hidden factors, take the U.S. lending club as an example, choose the lender information data of the whole year of 2019, and establish the P2P online credit default Logistic Regression prediction model. Considering that the conditional factors contain multiple value states, the One-Hot idea is introduced to improve the precision of the algorithm. The accuracy, recall and other evaluation indexes are chosen to compare and analyse the prediction effect of the model. The results of the model show that Logistic Regression can effectively predict the credit default risk of personal credit, and also provide a more in-depth explanation for the generation of personal credit default risk in the context of new personal loans.

Insurance companies always use federated learning to integrate external data sources for data analysis and improve the conversion rate of insurance precision marketing. However, due to imbalanced data distribution and the presence of null data, the joint modeling often suffers from low robustness and is prone to falling into the dilemma of under-fitting. Therefore, the feature selection for federated learning needs to be incorporated before the joint modeling to improve the accuracy of predictions. In this paper, we propose the FedPV-FS method, which includes two-party feature selection based on public verifiable covert (PVC), and multi-party federated feature selection based on verifiable secret sharing (VSS). Moreover, we iteratively optimize federated feature selection using data selection, transformation, and integration. Experiments show that our method can achieve high-quality feature selection for increasing the optimization objective to 88.4%, promote the continuous increase of insurance premiums, and has good applications in insurance precision marketing scenarios.

The protection of database watermarking techniques based on primary key (PK) is weakened by potential PK attacks and their huge impact. Using virtual primary key (VPK) schemes is an effective solution to enhance the robustness of these techniques. Pérez Gort et al. proposed the HQR-Scheme, which first considers balancing the participation rate of each attribute to improve the ability of scheme to cope with deletion attacks. However, the HQR-Scheme has limited control ability and cannot adapt to changes in relevant information in databases. Facing these challenges, we innovatively propose the FRBBM-Scheme, which allows users to fully analyze data table information and formulate flexible ratio adjustment strategies to enhance their ability to cope with deletion attacks. We verify our proposed scheme through multiple experiments, which show that it has excellent ratio control ability, can generate high-quality VPK sets, and can resist different levels of attribute deletion attacks.

Intrusion detection is a growing area of concern in Industrial Internet of Things (IIoT) systems. This is largely due to the fact that IIoT systems are typically used to augment the operation of Critical Information Infrastructures, the compromise of which could result in severe consequences for industries or even nations. In addition, IIoT is a relatively new technological development which introduces new vulnerabilities. Machine learning methods are increasingly being applied to IIoT intrusion detection. However, the data imbalance prevalent in IIoT intrusion detection datasets can limit the performance of intrusion detection algorithms due to the significantly smaller amount of attack samples. As such, generative models have been applied to address the data imbalance problem by modelling distributions of intrusion detection datasets in order to generate synthetic attack samples. Current work presents the implementation of a Generative Adversarial Artificial Immune Network (GAAINet) as an approach for addressing data imbalance IIoT intrusion detection. Experimental results show that GAAINet could generate synthetic attack samples for the WUSTL-IIoT-2021 dataset. The resulting balanced dataset was used to train an Artificial Immune Network classifier, which achieved a detection accuracy of 99.13% for binary classification and 98.87% for multi-class classification.

The Internet of Vehicles (IoV) improves efficiency of transportation systems while enhancing the passenger travel experience. However, due to the open wireless communication environment, the IoV requires a reliable and secure authentication and key agreement scheme to ensure that the exchanged data in public channel cannot be forged or modified by the adversary. In most existing authentication schemes, the vehicle usually authenticates with each other by an online Trusted Authority (TA), which results in the authentication efficiency of these centralized authentication schemes are easily affected by TA’s computational and communication bottlenecks as the increase of traffic density. Therefore, this paper proposes a secure and efficient authentication and key agreement scheme for IoV, where the vehicles can authenticate with each other and build a session key through a pre-shared key. Besides, a group key is also proposed to broadcast basic safety messages in the same RSU group securely. The group key can be updated when the vehicle joins and leaves, so that a leaving group member cannot access the current communication process. By the Heuristic and BAN logic analysis, the proposed scheme is proved to be secure. Compared with existing schemes, the proposed scheme meets the security requirements and has significant advantages in terms of computation and communication overhead.

In order to extract deeper features from surface electromyography signals and improve the classification accuracy of lower limb movements, a feature extraction method combining wavelet packet and sample entropy (WPT-SampEn) is proposed to accurately identify three types of lower limb movements. The electromyographic signals are preprocessed, which includes Butterworth filtering, activity segment detection based on short-term energy, and normalization processing. A three-layer wavelet packet decomposition method is used to decompose the five electromyographic signals into eight different frequency bands. By calculating the energy proportion in each frequency band, the top four frequency bands are determined as the focus of analysis. The Kruskal-Wallis test is employed to select frequency bands with statistical differences. To validate the effectiveness of this method, the support vector machine (SVM) algorithm is used for lower limb motion classification. Experimental results show that using the wavelet packet sample entropy features of the lateral thigh, medial thigh, rectus femoris, and biceps femoris muscles, the recognition rate reaches up to 96.46%. Compared with existing methods, this approach can extract deeper features from sEMG signals and achieve higher recognition accuracy. It has great potential in areas such as rehabilitation training, wearable exoskeleton control, and daily activity monitoring.