Complexity最新文献

Demand forecasting for new products in the fashion industry has always been challenging due to changing trends, longer lead times, seasonal shifts, and the proliferation of products. Accurate demand forecasting requires a thorough understanding of consumer preferences. This research suggests a model based on machine learning to analyse customer preferences and forecast the demand for new products. To understand customer preferences, the fitting room data are analysed, and customer profiles are created. K-means clustering, an unsupervised machine learning algorithm, is applied to form clusters by grouping similar profiles. The clusters were assigned weights related to the percentage of product in each cluster. Following the clustering process, a decision tree classification model is used to classify the new product into one of the predefined clusters to predict demand for the new product. This demand forecasting approach will enable retailers to stock products that align with customer preferences, thereby minimising excess inventory.

This paper develops two numerical methods for solving a system of fractional differential equations based on hybrid shifted orthonormal Bernstein polynomials with generalized block-pulse functions (HSOBBPFs) and hybrid shifted orthonormal Legendre polynomials with generalized block-pulse functions (HSOLBPFs). Using these hybrid bases and the operational matrices method, the system of fractional differential equations is reduced to a system of algebraic equations. Error analysis is performed and some simulation examples are provided to demonstrate the efficacy of the proposed techniques. The numerical results of the proposed methods are compared to those of the existing numerical methods. These approaches are distinguished by their ability to work on the wide interval [0, a], as well as their high accuracy and rapid convergence, demonstrating the utility of the proposed approaches over other numerical methods.

Back electromotive force (EMF)-based sliding mode observer (SMO) is increasingly employed for interior permanent magnet synchronous machine (IPMSM) sensorless drives due to its high robustness to external disturbance and low sensitivity to system parameter variations. However, its control performance is severely weakened by the inherent chattering and speed iteration operation. In order to effectively resolve these problems, a strategy to design a dual-SMO is proposed in this paper. With the proposed strategy, the combination of the stator-voltage transformation matrix (SVTM) and the low-pass filter is developed to obtain the rotor position information, which greatly alleviates the chattering without any deviations. Meanwhile, three independent equations are constructed and extracted by placing two SVTMs in different locations. By solving these three equations, the rotor position can be calculated directly with zero phase shift, which eliminates the speed iteration operation and improves the system’s dynamic performance. Furthermore, by analyzing the influences of machine parameters’ variations, the suitable virtual q-axis inductance can be selected to quickly achieve the optimal-efficiency sensorless control of the IPMSM. Finally, the experimental results on an IPMSM demonstrate that the rotor position with good steady-state and dynamic performance can be obtained accurately by using the proposed sensorless control strategy.

Problem Statement. Proper and timely relief in the postdisaster phase is very important to minimize victims and casualties. It is necessary to assess the needs of the affected points and provide relief in the shortest possible time without wasting time. For this purpose, it is necessary and vital to determine the location of care centers, temporary accommodation centers, and routing to distribution vital and medical items. The Proposed Approach. In this paper, using a scenario-based stochastic planning mathematical model, postdisaster relief is discussed. Also, attention has been paid to the distribution of vital items by using routing. Contributions. By using the epsilon constraint method, a strong efficient solution has been achieved for model’s objectives, and also a sensitivity analysis has been performed on some of the model’s parameters. Also, an accelerated stochastic benders decomposition algorithm is suggested to solve the problem modeled in this paper. To speed up the convergence of the solution algorithm, valid inequalities are introduced to get better quality lower bounds. Results. The results of the research show that the simultaneous consideration of the relief evacuation and distribution process improves the relief logistics process.

Analyzing the supply chain (SC) of biopharmaceutical drugs can be challenging due to their complexity, the existence of a wide variety of risks, and the dynamics of the system. This paper presents a framework for evaluating the SC risks of Iranian biopharmaceutical companies based on cause-and-effect relationships and fuzzy cognitive maps (FCMs). We first interviewed several biopharmaceutical supply experts to learn about potential SC risks, causal relationships among FCM concepts, FCM structure, and FCM activation cycle. The most critical and relevant risks and significant elements of the SCs, such as cost, time, and quality, were identified as relevant FCM concepts. Then, we used failure mode and effects analysis (FMEA) and the FCM of the SC risks to assess the impacts of the biopharmaceutical SC risks on each other and on the crucial elements of the SCs. The Hebbian learning algorithms were then applied to train the FCM models. We tested different scenarios to evaluate the impacts of FCM concepts on the SC risks. The proposed approach can prioritize risk factors and, more importantly, predict and analyze the effect of each risk factor/risk group on other risks or the outcome of a given risk. The proposed FCM features and the knowledge gained from evaluating them can provide practical and helpful information to pharmaceutical companies to deal with their supply risks more efficiently.

This article proposed an image encryption scheme along a novel five-dimensional hyperchaotic system and a Fibonacci Q-matrix (FQ-matrix) for gray images. This designed algorithm follows two key stages: the confusion stage and the diffusion stage. In the confusion step and diffusion step, the placement of the plain image pixels is replaced by a 5D hyperchaotic map and pixel values are changed using the FQ-matrix, respectively. The fortitude of the designed encryption scheme is concluded by countering the algorithm on statistical analysis like histogram analysis, chi-square test, correlation coefficient analysis, information entropy analysis, differential attacks (NPCR, UACI), and NIST suite test. The produced algorithm’s experimental evaluation indicates that the entropy, NPCR, and UACI values tend to be ideal values. The numerous analyses indicate that the proposed algorithm has a lot of characteristics like the low correlation of adjacent cipher pixels, strong security, and large key space, which can give high confidentiality in image data.

The periodic oscillation transmission of infectious diseases is widespread, deep understanding of this periodic pattern and exploring the generation mechanism, and identifying the specific factors that lead to such periodic outbreaks, which are of very importanceto predict and control the spread of infectious diseases. In this study, to further reveal the mathematical mechanism of spontaneous generation of periodic oscillation solution, we investigate a type of SEIR epidemic model with a small intrinsic growth rate. By utilizing the singular perturbation theory and center manifold theorem, we extend the relaxation oscillation of three-dimensional SIR models to the four-dimensional SEIR models and prove the existence of stable relaxation oscillation with a large amplitude in the model. Numerical simulations are performed to verify our theoretical results. The results presented in this study provide a new idea for the study of the intrinsic mechanism of periodic oscillation in epidemiology, enrich the dynamics of epidemic models, and deepen the understanding of the global dynamics of these models.

The present paper introduces a novel object of study, a language fractal structure; we hypothesize that a set of embeddings of all n-grams of a natural language constitutes a representative sample of this fractal set. (We use the term Hailonakea to refer to the sum total of all language fractal structures, over all n). The paper estimates intrinsic (genuine) dimensions of language fractal structures for the Russian and English languages. To this end, we employ methods based on (1) topological data analysis and (2) a minimum spanning tree of a data graph for a cloud of points considered (Steele theorem). For both languages, for all n, the intrinsic dimensions appear to be noninteger values (typical for fractal sets), close to 9 for both of the Russian and English language.

At present, there are three main methods for analyzing the causes of ship collision accidents: statistical analysis, accident causation models, and knowledge graphs. With the deepening of research, the analysis methods pay more attention to the objective correlation between various factors of the accident, and the analysis results obtained are more objective and accurate. On this basis, this paper proposes a method for analyzing the contribution degree of different causes and accident conduction paths in ship collision accidents based on the construction of the Ship Collision Accidents Event Graph (SCAEG). Firstly, the ontology is constructed based on the grounded theory. Secondly, events and relationships are extracted after fine-tuning the UIE model. Thirdly, the SCAEG is constructed after event coreference resolution. Finally, this research conducts the contribution degree analysis, accident conduction path analysis, and accident spatial distribution analysis based on SCAEG. The advantages of this method include the following: (i) it can construct a more complete and accurate ontology; (ii) adopting this approach can unify various information extraction tasks and achieve good results based on small sample annotation data; and (iii) using this method, we can conduct contribution degree analysis of different causes, accident conduction path analysis, and spatial distribution analysis. Experimental evidence demonstrates the effectiveness of this method. The analytical results obtained from the experiments can provide assistant decision-making for relevant departments to reduce the occurrence of ship collision accidents and improve maritime traffic safety.

In this paper, we analyze the problem of managing users from different slices connecting to a software-defined network (SDN). We seek to minimize the propagation latency between switches and controllers as well as between controllers themselves. We also minimize the connection latency between users and their network access nodes. Thus, the main highlights of the paper are to formally represent the problem utilizing two equivalent mixed-integer quadratic programming models. The first one represents the user requirements of each slice by using a membership matrix. The second one consists of subsets of users separated within each slice requirement. Subsequently, the above models are analyzed in a standard linearized version. Finally, they are compared with a proposed local search math-heuristic algorithm. The proposed models and algorithm are solved with the CPLEX solver with default options. To the best of our knowledge, this journal paper constitutes a first attempt to incorporate network slicing in SDN allowing flexibility, resource efficiency, security, and effective management of the network facilitating the deployment of customized and adaptive services. Besides, our models allow us to deal with the management of connecting users to either controller or switch-type nodes depending on the slice to which each user belongs. For security reasons, a certain slice could only have access to the network controllers, while the rest of the users that belong to the other slices can connect to the switch-type nodes of the network. From the numerical experiments, we observe that the linear models show a better performance in terms of CPU times and the best solutions obtained. Similarly, our proposed approximation algorithm achieves near-optimal solutions in significantly shorter CPU times, for all the input graph networks, when compared to the proposed exact models which allows for finding the optimal solutions.