Soft Computing最新文献

In recent years, the method of using graph neural networks (GNN) to learn users’ social influence has been widely applied to social recommendation and has shown effectiveness, but several important challenges have not been well addressed: (i) Most work fails to consider the user interests (historical user-item interactions) when first building user-user social relationships, which can make it difficult to capture accurate user embedding and thus prevent the model from better exploring the users’ social influence; (ii) Most of the current methods do not build social neighbors (with the same item-user interaction) that belong to the item and do not aggregate information from the perspective of social neighbors, which makes it possible for the item to lose a lot of details when expressing the user’s interest factors. Therefore, to address the above challenges, we propose Exploring Implicit Influence for Social Recommendation Based on GNN (EIIGNN). First, we construct the initial user embedding with user-item interaction information and use the implicit modeling module in user modeling to explore the implicit influence of interest factors on users. In addition, EIIGNN models the social graph structure of item (an item-item graph) so that item can aggregate information from the perspective of their social neighbors, which helps the model learn a more accurate representation of the item. Finally, extensive experimental results on two real-world datasets clearly demonstrate the effectiveness of EIIGNN.

This paper presents a review on methods for class-imbalanced learning with the Support Vector Machine (SVM) and its variants. We first explain the structure of SVM and its variants and discuss their inefficiency in learning with class-imbalanced data sets. We introduce a hierarchical categorization of SVM-based models with respect to class-imbalanced learning. Specifically, we categorize SVM-based models into re-sampling, algorithmic, and fusion methods, and discuss the principles of the representative models in each category. In addition, we conduct a series of empirical evaluations to compare the performances of various representative SVM-based models in each category using benchmark imbalanced data sets, ranging from low to high imbalanced ratios. Our findings reveal that while algorithmic methods are less time-consuming owing to no data pre-processing requirements, fusion methods, which combine both re-sampling and algorithmic approaches, generally perform the best, but with a higher computational load. A discussion on research gaps and future research directions is provided.

Most of the current attribute reduction methods use the measure to define the reduct, such as the positive region of rough set theory (RS), information entropy, and distance. However, the size of the reduct based on the measures is still limited. To cope with this problem, we propose a new approach of attribute reduction based on using the intuitionistic fuzzy topology (IFT). Firstly, a new IFT structure based on the pre-order relation and the intuitionistic fuzzy base (IF-base) structure is introduced. Secondly, a new measure is proposed to evaluate the significance of the attribute based on the IF subbase. Finally, the new reduction algorithms based on the IF-base filter and filter-wrapper methods are presented. The theoretical and experimental results show that the proposed method is efficient in terms of size and accuracy of the reduct. Specifically, the reduct of the F_IFT algorithm has an average size of 50% smaller, and the FW_IFT algorithm has an average accuracy of 10% greater than those of the related algorithms. Significantly, the algorithm FW_IFT can very remove noisy attributes. The classification accuracy of the reduct is 15% higher than that of the original set of attributes.

Classification predictive modeling involves the accurate assignment of observations in a dataset to target classes or categories. Real-world classification problems with severely imbalanced class distributions have increased substantially in recent years. In such cases, significantly fewer observations are available for minority classes to learn from than for majority classes. Despite this sparsity, the minority class is often considered as the more interesting class, yet the development of a scientific learning algorithm that is suitable for these observations presents numerous challenges. In this study, we further explore the merits of an effective multi-class classification algorithm known as SAMME.C2 that is specialized for handling severely imbalanced classes. This innovative method blends the flexible mechanics of the boosting techniques from the SAMME algorithm, which is a multi-class classifier, and the Ada.C2 algorithm, which is a cost-sensitive binary classifier that is designed to address highly imbalanced classes. We establish a scientific and statistical formulation of the SAMME.C2 algorithm, together with providing and explaining the resulting procedure. We demonstrate the consistently superior performance of this algorithm through numerical experiments as well as empirical studies.

Community detection is a valuable tool for studying the function and dynamic structure of most real-world networks. Existing techniques either concentrate on the network's topological structure or node properties without adequately addressing the dynamic aspect. As a result, in this research, we present a unique technique called Multi-Objective Optimization Overlapping Dynamic Community Detection (MOOODCD) that leverages both the topological structure and node attributes of dynamic networks. By incorporating the Dirichlet distribution to control network dynamics, we formulate dynamic community detection as a non-negative matrix factorization problem. The block coordinate ascent method is used to estimate the latent elements of the model. Our experiments on artificial and real networks indicate that MOOODCD detects overlapping communities in dynamic networks with acceptable precision and scalability.

In various applied disciplines, the modeling of continuous data often requires the use of flexible continuous distributions. Meeting this demand calls for the introduction of new continuous distributions that possess desirable characteristics. This paper introduces a new continuous distribution. Several estimators for estimating the unknown parameters of the new distribution are discussed and their efficiency is assessed through Monte Carlo simulations. Furthermore, the process capability index (S_{pmk}) is examined when the underlying distribution is the proposed distribution. The maximum likelihood estimation of the (S_{pmk}) is also studied. The asymptotic confidence interval is also constructed for (S_{pmk}). The simulation results indicate that estimators for both the unknown parameters of the new distribution and the (S_{pmk}) provide reasonable results. Some practical analyses are also performed on both the new distribution and the (S_{pmk}). The results of the conducted data analysis indicate that the new distribution yields effective outcomes in modeling lifetime data in the literature. Similarly, the data analyses performed for (S_{pmk}) illustrate that the new distribution can be utilized for process capability indices by quality controllers.

Workload migration among cloud data centers is currently an evolving task that requires substantial advancements. The incorporation of fuzzy systems holds potential for enhancing performance and efficiency within cloud computing. This study addresses a multi-objective problem wherein the goal is to maximize the interpretability and the percentage of renewable energy consumed by a fuzzy meta-scheduler system in cloud scenarios. To accomplish this objective, the present research proposes a novel approach utilizing a multi-objective Knowledge Acquisition with a Swarm Intelligence Approach algorithm. Additionally, it takes advantage of a framework built on CloudSim, which includes virtual machine migration capabilities based on an expert system. Furthermore, a hierarchical fuzzy system is employed to assess rule base interpretability, along with another multi-objective algorithm, named Non-dominated Sorting Genetic Algorithm II. The framework and hierarchical system are employed to perform various simulation results concerning renewable energy and interpretability, while the algorithms aim to enhance the system’s performance and interpretability. Empirical results demonstrate that it is possible to improve the performance of cloud data centers while improving the interpretability of the corresponding fuzzy rule-based system. The proposed multi-objective algorithm shows comparable or superior performance to the genetic algorithm across diverse scenarios. The simulation results indicate that improvements in cloud data center performance can be achieved while enhancing system interpretability. The average improvement in the interpretability index ranges from 0.6 to 6%, with a corresponding increase in renewable energy utilization ranging from 5 to 6%.

Learning high order non-oscillatory polynomial approximation procedures which form the backbone of high order numerical solution of partial differential equations is challenging. The major issue is to pose these procedures as a learning problem and generate suitable synthetic data set which suffice learning it with small neural networks. In this work, we pose an arc-length based essentially non-oscillatory (ENOL) reconstruction algorithm as machine learning problem. A novel way to construct the synthetic data using ENOL algorithm along with basic smooth and piece-wise continuous functions is given. Small vanilla regression and classification neural networks are trained to learn third order (ENOL) polynomial approximation procedure. The metric of trained ENO classification and regression networks is presented and commented. These trained models are implemented in numerical solver to compute the solution of test problems of hyperbolic conservation laws. The presented numerical results show that ENO classification network gives results comparable to the exact ENOL reconstruction whereas ENO regression network performs poorly both in terms of convergence and resolving the discontinuities.

This study uses the cubic spline method to solve the one-dimensional (1D) (one spatial and one temporal dimension) heat problem (a parametric linear partial differential equation) numerically using both explicit and implicit strategies. The set of simultaneous equations acquired in both the explicit and implicit method may be solved using the Thomas algorithm from the tridiagonal dominating matrix, and the spline offers a continuous solution. The results are implemented with very fine meshes and with relatively small-time steps. Using mesh refinement, it was possible to find better temperature distribution in the thin bar. Five numerical examples are used to support the efficiency and accuracy of the current scheme. The findings are also compared with analytical results and other results in terms of error and error norms ({L}_{2}) and ({L}_{infty }). The Von-Neuman technique is used to analyse stability. The truncation error of both systems is calculated and determined to have a convergence of order (Oleft( {h + Delta t^{2} } right).)

Optimization problem involving interval parameters and multiple conflicting objectives are called multi-objective optimization problems with interval parameters (IMOPs), which are common and hard to be solved effectively in practical applications. An interval multi-objective honey badger algorithm (IMOHBA) is proposed to address the IMOPs in this paper. Firstly, the (mu) metric is employed to assess the Pareto dominance relationship among interval individuals, which reflects the quality of the optimal solutions. Secondly, the crowding distance suitable for the interval objective is utilized to reflect the distribution of the optimal solution. Finally, the candidate solutions are ranked and selected by the non-dominated sorting method. To validate the performance of IMOHBA, it is tested on 19 benchmark IMOPs as well as an interval multi-objective scheduling problem for underwater wireless sensor networks and compared with three state-of-the-art algorithms. The experimental results demonstrate the superiority and strong competitiveness of IMOHBA in addressing IMOPs, exhibiting improved convergence and broader exploration capabilities of the solution space. These findings further validate the effectiveness and feasibility of IMOHBA, highlighting its unique advantage in solving IMOPs.