Soft Computing最新文献

Particle swarm optimization (PSO) is a classical computational method that optimizes a problem by iteratively trying to find the optimal solution. It still suffers somes defects such as poor local search ability, low search accuracy and premature convergence, especially in high-dimensional complex problems. In order to address these issues, this paper has proposed a novel PSO algorithm with distributed delays of adaptive weights and evolved acceleration coefficients (PSO-DWC). The main idea of the proposed improved PSO algorithm is three-fold: (1) a mechanism is introduced to evaluate the current evolutionary state by evolutionary factors of the swarm and to predict the next state by a probability transition matrix; (2) distributed time-varying time-delays are added into the velocity updated model; (3) adaptive inertia weight varies according to evolutionary factors, which describes the population distribution information; and newly-introduced evolved acceleration coefficients are determined by the predict next evolutionary state of the swarm. Owing to the promising issues mentioned above, the PSO-DWC algorithm has the advantages of keeping the diversity of particles, balancing the local and global search abilities and reaching to an acceptable solution. Experiments on twenty well-known benchmark functions have demonstrated that the proposed PSO-DWC algorithm has a superior performance over other five well-known PSO algorithms in high dimensional search space. Statistical significance tests verify the superiority of the new algorithm. Therefore it can be concluded that the novel PSO-DWC algorithm is able to solve the optimization problems with powerful global search and efficient convergence.

The kNN classifier is the most popular, supervised machine-learning technique, but the main disadvantage of this algorithm is that it has restricted access to the class distributions in a training point set and treats all the instances equally. In kNN classification, fuzzy sets are used to obtain the membership degrees of each point to the classes known as fuzzy kNN (FkNN) classification. Although the FkNN classifier enhances the performance of the kNN, it does not consider the effect of noisy and redundant instances, which makes it ineffective. Moreover, the performance of kNN is dependent on the value of k (number of nearest neighbours). Considering these issues, we present a novel algorithm that simultaneously tunes the class-dependent feature weights and k value using an effective meta-heuristic algorithm, the Enhanced Equilibrium Optimization technique. Several experiments have been conducted on different biomedical datasets, and the proposed approach has outperformed the other standard classifiers in terms of accuracy.

Short-term load forecasting (STLF) usually begins by grouping data according to various criteria, most often by days of the week. Then, based on the obtained segments, independent models are created. Each model’s prediction uses only one segment of the data. This paper proposes a new approach to model formation based on the correlation between the forecasted day and previous days. The proposed approach is compared with the usual approach where data segments are obtained by grouping according to days of the week. The models were created using fuzzy logic and ANFIS. The mean absolute percentage errors of the new approach and the usual approach using ANFIS in terms of prediction accuracy are obtained as 2.89 and 4.15, respectively. The mean absolute percentage errors for the new approach and the usual approach are 3.39 and 4.78, respectively, when fuzzy logic is used. The results showed that when the proposed method is used, forecasts for the day ahead are much more accurate in both cases.

Twin Support Vector Machine (TSVM) transforms a single large quadratic programming problem (QPP) in support vector machine (SVM) into two smaller QPPs by finding two non-parallel classification hyperplanes, so that its computational time is reduced to a quarter of what the traditional SVM takes. However, TSVM ignores the data distribution of class, which makes TSVM sensitive to noise. In this paper, a fuzzy twin support vector machine based on centered kernel alignment (FTSVM-CKA) is proposed to solve the problem that TSVM is sensitive to noise. Firstly, a feature-weighted kernel function is constructed by using the information gain, and it is applied to the calculation of the centered kernel alignment (CKA). This assigns greater weight to strongly correlated features, emphasizing their classification importance over weakly correlated features. Secondly, the CKA method is utilized to derive a heuristic function for calculating the dependency between samples and their corresponding labels, which assigns fuzzy membership to different samples. Based on this, a fuzzy membership assignment strategy is proposed that can effectively address the sensitivity of TSVM to noise. Thirdly, this strategy is combined with TSVM to propose the FTSVM-CKA model. Moreover, this study employs a coordinate descent strategy with shrinking by active set to tackle the computational complexity arising from high-dimensional inputs. This can effectively accelerate the training speed of the model while ensuring classification performance. In order to evaluate the performance of FTSVM-CKA, this study conducts experiments designed on artificial and UCI datasets. The results demonstrate that FTSVM-CKA can efficiently and quickly solve binary classification problems with noise.

The substitution box (S-Box) plays a fundamental role in cryptographic algorithms. In this article, the Lu-Chen system is used to design a chaotic S-Box. We design two S-Boxes, one based on the rotation algorithm relative to the rows (or columns) and the other based on the Zigzag transformation. The performance of the new S-Boxes is evaluated with the bijective, nonlinearity, strict avalanche criterion, output bit independence criterion, differential approximation probability, linear approximation probability, algebraic degree and not having a fixed point and opposite fixed point. The analysis results show that the proposed S-Boxes have suitable cryptographic properties. In addition, an image encryption algorithm based on two generated S-Boxed, and a generalized Lai–Massey structure is presented. Experimental results show that the proposed method has achieved acceptable security.

Several factors affect customers' decisions regarding service selection. Two of the essential factors are the price and the service quality. The seller's credibility is also among the influential factors in selecting a service that is reinforced by advertising. The emergence of the Internet has led to increasing attention by sellers to advertising through online group buying (OGB) platforms. Sellers aim to attract new customers by offering discounts on OGB platforms. In this paper, the seller can sell its service through offline and online channels during two current and future courses. The aim is to determine the optimal strategy for the seller when deciding to join the OGB platform. All parameters of the problem are determined as fuzzy variables. Accordingly, this paper develops a fuzzy mathematical model to simultaneously determine price, service quality, and advertising level in a dual channel supply chain. A cooperative game between the seller and the OGB platform is developed under different refunding and revenue sharing scenarios for the centralized model. The optimal solutions of the problem are then defined using the game and fuzzy sets theories for each scenario. A numerical example is presented to indicate the effectiveness of the theoretical results of the models and developing management insights. In addition, sensitivity analyses also provide the effect of changes in essential parameters on the seller’s decisions.

Fake news creates social turbulence, which may hamper our social or economic equilibrium. Researchers have harnessed machine learning (ML) and deep learning (DL) algorithms to combat this challenge, particularly in disparate environments. Numerous techniques have been created to classify false news based on various textual features, including deep learning, machine learning, and evolutionary methodologies. Although fake news sentiment analysis is not entirely new, sentiment score-based artificial news analysis is rarely used. Our method incorporates machine learning techniques and deep learning techniques, such as LSTM-BiLSTM, with SentiWordNet parser-obtained sentiment scores. This integration improves feature sets and enables a more detailed analysis of emotional context. This research pioneers using machine learning along with deep learning techniques based on sentiment scores, an innovative approach within the field. Our research substantially improves the detection of false news. Recall and F-measure are significantly enhanced using machine learning techniques with the COVID-19 dataset. Moreover, sentiment-based deep learning techniques used for both the LIAR and COVID-19 datasets surpass previous benchmarks, obtaining a remarkable accuracy improvement of over 15% on the LIAR dataset compared to existing literature. This pioneering sentiment score-based approach enhances fake news detection accuracy, offering a potent tool to counter misinformation and safeguard societal equilibrium.

This paper proposes a new design method of energy management strategy (EMS) with adaptive super-twisting sliding mode control (ASTSMC) for fuel cell/battery/supercapacitor hybrid vehicle (FCHEV). The main objective of the proposed EMS is to improve power performance, fuel cell lifetime, and fuel consumption while considering the regulation of the DC-bus voltage. The proposed EMS is designed based on a frequency-decoupling technique using an adaptive low-pass filter, Harr wavelet transform (HWT), and FLC to decouple the required power into low, medium, and high-frequency components for fuel cell, battery, and supercapacitor, respectively. The presented frequency-decoupling-based strategy can improve the power performance of the vehicle as well as reduce load stress and power fluctuation in the fuel cell. Nevertheless, the neural network optimization algorithm (NNOA) is employed to optimize the membership functions of FLCs while considering the hydrogen consumption and constraints on the state of charge (SOC) of the battery and supercapacitor. To achieve robustness and high precision control, the ASTSMC is developed based on a nonlinear disturbance observer (NDOB) to stabilize the DC-bus voltage and currents of the energy sources, ensuring that the fuel cell, battery, and supercapacitor track their obtained reference values. The FCHEV system with the proposed EMS is modeled on MATLAB/Simulink, and three typical driving cycles such as HWFET, UDDS, and WLTP driving schedules are used for evaluation. The findings exhibit that the proposed EMS can effectively improve the fuel economy, reduce power fluctuation in the fuel cell, and prolong its lifetime compared to other existing methods such as the equivalent consumption minimization strategy (ECMS), state machine (SM), and FLC-based EMSs.

Sustainable supplier selection and optimal quantity transportation (S(^3)OQT) play an important role in supply chain management. This research represents a new four-stage solution approach for (hbox {S}^3)OQT where the multi-criteria decision making (MCDM) methods are integrated through an optimization model. In first stage, a new uncertainty interval type-2 spherical fuzzy set (IT2SFS) is introduced to help the decision-makers (DMs) for securing and reliable results in hesitancy situations. We develop a new operator on IT2SFS under Dombi t-norm and t-conorm by integrating Muirhead mean (MM) operator based on Choquet integral (CI). The preferences and priorities to the sustainable criteria based on interaction and interrelationship are represented by CI. Thereafter, the weights of the criteria and sub-criteria are determined by CI-indifference threshold-based attribute ratio analysis (ITARA) method by utilizing the proposed operator. In second stage, to evaluate the weights of the suppliers and to rank these, we construct a new MCDM method CI-TODIM (an acronym in Portuguese of interactive multi-criteria decision-making)-measurement alternatives and ranking according to compromise solution (MARCOS) method by utilizing the proposed operator and then finally design a new ranking function. In third stage, a new model on stochastic multi-objective mixed-integer non-linear solid transportation problem ((hbox {SM}^2)NSTP) is established to identify suitable supplier under sustainable risk criteria, and then, optimal quantity of products are transported from each supplier. Thereafter, we propose TOPSIS-neutrosophic-game theoretic approach (TNGTA) to obtain Pareto-optimal solution. We apply (varepsilon )-constraint method to obtain Pareto-optimal solution from (hbox {SM}^2)NSTP model. In the fourth stage, a comparative study is drawn among the obtained Pareto-optimal solutions that are extracted from TNGTA and (varepsilon )-constraint method. Finally, two MCDM models, CRITIC-TOPSIS and CRITIC-MARCOS, are used to help the DMs for selecting the final Pareto-optimal solution. A real-life example is included to show the applicability and effectiveness of the designed hybrid MCDM-(hbox {SM}^2)NSTP model.

The Internet of Things (IoT) is a network of devices that can communicate and cooperate over the Internet. As the IoT expands, guaranteeing the dependability and accuracy of communication systems becomes increasingly important. One of the key challenges faced in the process of system development is the need to detection the errors in the early phases of system development. Formal techniques are the gold standard for ensuring a system’s correctness. In the context of the IoT, this paper presents an Event-B formal model for the verification of the correctness of Content-Based Publish/Subscribe Systems (CBPS). We developed our model using Event-B, which is an incrementally formal technique with a plugin-supported platform. Furthermore, it supports both theorem proving and model checking. The incremental method uses a series of refining processes to help manage complexity. The paper offers a thorough exposition of the CBPS architecture, with an emphasis on decentralised design, reliable message delivery, and message ordering. This formalised method ensures that the CBPS system satisfies its criteria and free of errors. As a case study for our concept, we employ a smart home system. Finally, we validate and verify the formal model using proof obligations and the Rodin platform.