查看更多>>摘要：Numerous studies have adopted deep learning (DL) in financial market forecasting models owing to its superior performance. The DL models require as many relevant input variables as possible to improve performance because they learn high-level features from these inputs. However, as the number of input variables increases, the number of parameters also increases, leaving the model prone to overfitting. We propose a novel stock-market prediction framework (LSTM-Forest) integrating long short-term memory and random forest (RF) to address this issue. We also develop a multi-task model that predicts stock market returns and classifies return directions to improve predictability and profitability. Our model is interpretable because it can identify key variables using the variable importance analysis from RF. We used three global stock indices and 43 financial technical indicators to verify the proposed methods experimentally. The root mean squared errors of LSTM-Forest with multi-task (LFM) in predicting returns from S&P500, SSE, and KOSPI200 were 25.53%, 22.75%, and 16.29% lower, respectively, than those of the baseline RF model. The model's balanced accuracy in forecasting the daily return direction increased by 7.37, 1.68, and 3.79 points, respectively. Furthermore, our multi-task model outperforms our single-task model and previous DL approaches. In the trading test, LFM produced the highest profits-even compared with the long-only strategy when transaction costs were considered. The proposed framework is readily extensible to other tasks and fields that contend with high-dimensionality problems. (C) 2021 Elsevier B.V. All rights reserved.
查看更多>>摘要：We discuss the portfolio selection problems in which the uncertainty of future returns and the heterogeneity of investor attitudes towards the stock market (optimistic-pessimistic-neutral) are captured by coherent fuzzy numbers. Two coherent fuzzy multi-period portfolio selection models are developed from the perspectives of wealth maximization and risk minimization. Given that the constraint levels regarding risk and return of the current period tend to be influenced by the outcome of the previous period, the dynamic risk-tolerance and expected-return levels are integrated into the portfolio modeling. Practical constraints and transaction costs are also taken into account, which enables the models more effective and lifelike in simulating the real-world trading of the stock market. The empirical studies based on two large data sets are presented to illustrate the applicability of the proposed models. To survey the models' performance, several portfolio evaluation criteria are used to conduct out-of-sample analysis. The results show outstanding performance of the presented models with dynamic strategies over conventional ways (static risk-tolerance and expected-return levels) on most of the indicators. This research offers references for investors with different attitudes to make long-term investment decisions, and is an effective supplement to behavioral portfolio selection research based on bounded rationality under uncertainty. (C) 2021 Elsevier B.V. All rights reserved.
查看更多>>摘要：Blocking flow shop scheduling problem widely exists in industrial processes, and most attention has focused on minimization of economic indicators, such as makespan, total flow time, and due-date-based functions, rather than energy-efficient indicators. This paper considers idle and blocking time criterion, which is closely related to machine energy consumption in blocking flow shops, and proposes a migrating birds optimization with a diversified mechanism (dMBO) for the problem. On the basis of the profile fitting (PF) heuristic and the characteristics of the idle and blocking time, an improved heuristic, named PFI, is proposed by modifying the PF and performing an insert procedure. A best insert operator and an insert-based local search are hybridized in the proposed migrating birds optimization algorithm to enhance its exploitation capability. In order to maintain the diversity of the flock in the algorithm, a diversified mechanism containing three tabu lists and one candidate pool is designed. Extensive computational results validate the effectiveness of the proposed PFI heuristic, and a statistical analysis of the computational results confirms the superiority of the dMBO over several other high-performing metaheuristics. (C) 2021 Elsevier B.V. All rights reserved.
查看更多>>摘要：Cloud manufacturing (CMfg) has received increasingly attention from both academia and industry. Cloud service composition is a critical technique in CMfg that connects different available manufacturing cloud services (MCSs) to generate a composite manufacturing cloud service (CMCS) to satisfy users' requirements. Many available MCSs with the same or similar functionality but different QoS attributes are deployed in the CMfg platform. So it is challenging to obtain an optimal CMCS to satisfy the users' complex requirements. Considerable numbers of approaches have been proposed to solve this problem. However, most of them often fall in a local optimum instead of the global one. In this paper, a novel eagle strategy using uniform Mutation and modified Whale Optimization Algorithm (MWOA) is proposed to maintain a balance between the global and local search abilities. In this approach, the uniform mutation is applied to perform the global search to preserve the diversification of the population, and a modified whale optimization algorithm is designed to perform the local search. The performance of the new approach is verified on various benchmark functions and different scales of QoS-aware cloud service composition problems. The experimental results demonstrate that the proposed MWOA has superior performance over the other methods. (C) 2021 Published by Elsevier B.V.
查看更多>>摘要：The spectrum overcrowding is one of the prime issues faced by wireless telecommunication based applications. The network blockage causing the disconnection or call drops is another important concern. These problems, are needed to be addressed for implementing the 5G and beyond technologies. Therefore, to tackle the issues of spectrum overcrowding and network blockage simultaneously a Cognitive Radio (CR) technology based relay network is proposed in this work. The accurate detection of the primary user's signal by the cognitive radio users is the most integral functioning of the cognitive radio networks. The existing spectrum sensing using Deep Neural Network (DNN) and Convolutional Neural Network (CNN) techniques have their limitations concerned with accurate prediction and classification of vacant spectrum due to their tendency of getting jammed to the local optima. In this paper, we firstly propose a novel mutated Modified Whale Optimization Algorithm (MWOA) trained Spiking Neural Network (SNN) based spectrum sensing technique for the efficient detection of spectrum holes. Here, the weights of the SNN are trained by means of MWOA for efficiently predicting the spectrum holes. The proposed scheme exploits underlying structural information of the sensed signals via continuous wavelet transforms. The proposed scheme does not require any priori information about the channel state and is shown to achieve state of the art performance in the detection of spectrum holes. The simulation results have inferred that the proposed CR based relay model with the MWOA trained SNN based spectrum sensing has significantly improved the performance of the User Equipment (UE) in the network blockage area in terms of higher opportunistic throughput and lower BER (Bit Error Rate). The MWOA has proved to be an efficient training algorithm for SNN with the validation accuracy of 98%. (C) 2021 Elsevier B.V. All rights reserved.
查看更多>>摘要：Accurate hourly PM2.5 concentration prediction plays a key role in air quality monitoring and controlling system, especially when severe haze occurs frequently. A PM2.5 hourly prediction system is developed in this paper, based on an advanced data processing strategy, an effective feature selection technology and a novel optimization algorithm. First, the collected original sequence is decomposed into a group of filters with different wave frequencies and each filter is weighted and reconstructed to mitigate the negative impact of noisy fluctuations. Then mRMR is introduced for extracting interaction information between pollutants, further determining the input of artificial intelligence models. Whereafter, a five-component combined system is taken shape, in which BPNN, ELM, GRNN and BiLSTM are employed as foundation models while Multi-objective Water Cycle Algorithm (MOWCA) is the weight optimization model. The results of hourly PM2.5 concentration prediction simulation experiment in the Bei-Shang-Guang-Shen area make clear that the developed system with minimum forecasting error, excellent generalization capability and robust prediction performance shows a definite latent capacity and future to deal with early warning problems and to design suitable abatement strategies. (C) 2021 Elsevier B.V. All rights reserved.
查看更多>>摘要：Governmental limitations and customer expectations have increased the focus on sustainability in supply chain network design (SCND). To address this issue, a bi-objective mixed-integer mathematical model is introduced with the aim of simultaneously maximizing supply chain profit and minimizing overall carbon emissions. Additionally, customer demands are considered price- and greenness-sensitive for multiple products, and uncertainty in the production process is estimated by a finite number of scenarios. The Monte Carlo sampling approach is used to produce the initial scenarios, and a heuristic scenario reduction approach is subsequently utilized. Due to the complexity of the model, we develop a hybrid metaheuristic algorithm that embeds variable neighborhood search (VNS) in two genetic algorithms to accelerate the convergence of the algorithm to high-quality solutions. These algorithms are compared to multi-objective particle swarm optimization (MOPSO) with two leader selection procedures. To improve the performance of these algorithms, the response surface method is applied to modulate the algorithm parameters. Finally, several analyses are performed to investigate the efficiency and effectiveness of the proposed approach. (C) 2021 Published by Elsevier B.V.
查看更多>>摘要：The involvement of competition in supply networks has changed the existing monopoly platform in different fields. This paper examines a multi-level decision-making framework within a triple-stage strategic approach in competitive supply networks. The various levels of these supply networks consist of parent firms (parent brands), manufacturing plants, state-owned logistics company and franchised sales centers. The parent brands, while following the strategies of the state logistics company (as the leader of the game), seek to further expand their market share in the production, supply and sales sectors. The main contributions of the proposed approach are: the existence of partnership and non-partnership synergies in different stages of planning, the emergence and development of supply networks based on downstream alliances, the design of a multi-agent distribution mechanism based on environmental sustainability requirements, and the simultaneous development of cooperation and competition in terms of virtual alliances. Further, given the features of the issue under discussion, a hybrid Benders Decomposition-Particle Swarm Optimization algorithm is utilized. The designed structure of the algorithm helps to facilitate high-dimensional problem-solving while also addressing the interactive requirements of competitive games. The results of comparing the proposed solution approach with a game-theoretical heuristic, pure Benders decomposition and bi-level sub-population genetic algorithm prove its better performance, especially in large-size instances. (C) 2021 Elsevier B.V. All rights reserved.
NSTL
Elsevier