查看更多>>摘要：Engine faults, which are difficult to be accurately diagnosed, seriously affect the normal running of vehicles. To solve this problem, a novel fault diagnosis method via HOSVD–HOALS? hybrid algorithm-based multi-dimensional feature extraction was proposed for a vehicle engine. First, multiple source signals-based engine status samples with the form of third-order tensors were constructed to retain the correlation among sample data. Then, by analyzing the high-dimensional characteristics of the constructed tensor samples, the Tucker decomposition was employed to realize the dimension reduction of the samples. Simultaneously, combining the high order singular value decomposition (HOSVD) and high order alternation least square (HOALS), a hybrid algorithm was proposed to solve the optimal low-dimensional core tensors and features of the constructed tensor samples. Finally, based on the extracted features, the fault pattern of the vehicle engine was recognized by using the derived tensor-based K-nearest neighbor (K-NN) and fuzzy C-mean (FCM) algorithms, respectively. Research results show that the average accuracy of the fault diagnosis via HOSVD–HOALS hybrid algorithm-based multi-dimensional feature extraction can reach 96.25% which is the highest compared with that via the principal component analysis (PCA), HOSVD, HOALS, and direct tensor sample method, respectively, and the computation time required for the fault diagnosis is greatly shortened, which can provide theoretical and technical support for the fault diagnosis of vehicle engines.
查看更多>>摘要：The mainstream computing technology is not efficient in managing massive data and detecting network traffic intrusions, often including big data. The intrusions present in sustained network traffic and the massive host log event data cannot be effectively managed by conventional analytical tools, resulting in a huge number of false positives and a longer training time. This paper presents a novel technique to enhance the intrusion detection process by handling the fundamental big data complexities associated with different forms of heterogeneous security data. To achieve the earlier objective, the ensemble Support Vector Machine (SVM) is integrated with the Chaos Game Optimization (CGO) algorithm. The proposed methodology improves the intrusion classification accuracy and also identifies nine different types of attacks present in the UNSW-NB15 dataset. The efficiency of the proposed methodology is evaluated using statistical analysis and different performance metrics such as precision, recall, F1-score, accuracy, ROC curve, and confusion matrix by comparing it with different baseline models. The proposed methodology obtains an accuracy of 96.29% when compared to the chi-SVM (89.12%) and an improvement of 6.47% is noted in the proposed methodology in terms of accuracy when compared with the chi-SVM. The higher classification accuracy shows that the proposed methodology exhibit a fewer number of false positives when handling the security events in big data platforms.
查看更多>>摘要：The computation resource allocation is a key issue to the multiobjective evolutionary algorithms. Present studies still have some difficulties addressing this issue such as low accuracy, indispensable parameter for balancing the convergence and diversity, discontinuous improvement values of different domination relations and poor discernibility. To solve these problems, a hypervolume distribution model regarding two non-adjacent individuals of each subproblem is proposed in this paper. Based on the hypervolume distribution model, a comprehensive hypervolume distribution entropy (HDE) by integrating the entropy and the Kullback–Leibler divergence is proposed to measure the improvement of the subproblems. The proposed measurement is continuous over different domination relations, requires no parameters and has high precision and discernibility. Thereafter, a hypervolume distribution entropy guided multiobjective evolutionary based on decomposition (HDE-MOEA/D) is proposed. The proposed algorithm is more efficient on solving multiobjective optimization problems. The proposed algorithm is tested on some popular test suits against another five typical and popular algorithms. The proposed HDE-MOEA/D achieves the best generational distances and the inverse generational distances in 57.9% ranking comparisons. The HDE-MOEA/D also outperforms another compared algorithm in 70.2% one-to-one comparisons. The experiment results prove the superiority of the proposed algorithm and reveal some important discoveries.
查看更多>>摘要：The emergence and application of soft computing have significantly changed the methods to solve engineering problems. For the indoor environmental control, various machine learning methods have been used, attempting to replace the traditional engineering methods. However, the recent single data-driven machine learning approach can hardly meet the requirement, since engineering problems still require prior knowledge to extract features and set up restrictions, while this knowledge should be obtained from engineering analysis. In this case, this paper has proposed an integrated multi-discipline method combining machine learning with engineering analysis, to implement predictive intelligent indoor environmental control in terms of thermal comfort and energy consumption. This method includes three parts, i.e., environmental modelling and simulation, producing an environmental prediction model, and creating an intelligent control agent and system. Firstly, a physical model is created to simulate the indoor environment and analysed through computational fluid dynamics, whose results can guide the setup of sensors in the indoor environment for collecting real-time data. Then, a machine learning method support vector regression is used to create an environmental prediction model for key parameters within the indoor environment, based on the collected data. Finally, a reinforcement learning method is used to train an intelligent agent for the intelligent control on the indoor environment, together with a system for implementation. Experiments and evaluations are carried out in a case study within an office, demonstrating the proposed method's feasibility, which provides a more efficient and effective intelligently predictive control on the indoor environment considering the balance of thermal comfort and energy efficiency.
查看更多>>摘要：It is a grand challenge to detect anomalies existing in subspaces from a high-dimensional space. Most existing state-of-the-art methods implicitly or explicitly rely on distances. Since the contrast, e.g., distances, between data objects in a high-dimensional space becomes more and more similar. Moreover, high-dimensional spaces may include many irrelevant attributes masking anomalies (if the prior probability for a class remains unchanged regardless of the value observed for attribute att, att is said to be irrelevant to a class, i.e., att is an irrelevant attribute). Obviously, anomalies can exist in any of subspaces, so it is difficult to select subspaces that highlight the relevant attributes in an exponential searching space. To address this issue, we proposed a hybrid method consisting of a deep network and a hypersphere to detect anomalies. The deep network in the proposed method is used as a feature extractor to capture the low-dimensional features from the background space. Then, anomalies are separated by using the hypersphere in the feature space reconstructed by probability distribution. To prevent irrelevant attributes from being mistaken for anomalies during mining anomalies, the upper of the number of anomalies is estimated by the Chebyshev theorem. Finally, the proposed method was verified on synthetic datasets and real-world datasets. Experimental results show that the proposed method outperforms the existing state-of-the-art detection methods in regard to the accuracy of mining anomalies and the ability of noise resistance. We find that feature extractors can improve the ability of noise resistance for anomalous detection methods. In the feature space reconstructed by probability distribution, anomalous features are easily identified from irrelevant features and normal features. We also indicate that irrelevant attributes increase the complexity of the feature space, through calculating the probability distribution of data in the background space, the layered features can be extracted to distinguish anomaly classes, normal classes, and irrelevant attribute classes.
查看更多>>摘要：We address the problem of learning the parameters of the outranking-based multiple criteria sorting model from large sets of assignment examples. We focus on a recently devised method called Electre TRI-rC, incorporating a single characteristic profile to describe each decision class. We introduce four algorithms aimed at the problem. They use different optimization techniques, including an evolutionary algorithm, linear programming combined with a genetic approach, simulated annealing, and a dedicated heuristic. We present the results of the experiments carried out on both artificial and real-world data sets. They reveal an impact of the comparison and veto thresholds, various sorting rules, and ensembles on the classification accuracy of the proposed algorithms. From a broader perspective, we contribute to cross-fertilizing the fields of Multiple Criteria Decision Aiding and Machine Learning for supporting real-world decision-making.
查看更多>>摘要：The particle swarm optimization (PSO) algorithm has been widely used to solve optimization problems. One of its main drawbacks is a slow and sometimes premature convergence, which traps the swarm in a local minimum. Several PSO variants have been proposed to alleviate this phenomenon. Still, when dealing with structural damage detection (SDD), the performances of these algorithms are not homogeneous and highly depend on the number of defects and their severities, which even lead them to require assistance from other methods. This paper proposes a multi-component PSO with cooperative learning named MuC-PSO. Instead of resorting to other methods, a strategy pool is constructed in the proposed MuC-PSO by combining four PSO variants, which guarantees that the algorithm can execute multiple search strategies collaboratively. Moreover, a leader learning mechanism (LLM) is also implemented to ensure information exchange and lead the global convergence of the method. This strategy allows the PSO variants to benefit from each other and enables the MuC-PSO to solve complex SDD problems. The performances of the MuC-PSO are evaluated in nine damage scenarios of single and multiple damages with severity levels between [10%,50%]. Our algorithm is compared with different recent optimization algorithms, including the four PSO variants alone and some non-PSO algorithms. The simulation results on three types of damages clearly demonstrate the effectiveness and superiority of the MuC-PSO.
查看更多>>摘要：Echo state network (ESN) refers to a popular recurrent neural network with a largely and randomly generated reservoir for its rapid learning ability. However, it is difficult to design a reservoir that matches a specific task. To solve the structure design of the reservoir, a pseudo-inverse decomposition-based self-organizing modular echo state (PDSM-ESN) is proposed. PDSM-ESN is constructed by growing–pruning method, where the error and condition number are used, respectively. Since the self-organizing process may negatively affect the learning speed, the pseudo-inverse decomposition is adopted to improve learning speed, which means the output weights are learned by an iterative incremental method. Meanwhile, to solve the ill-posed problem, the modular sub-reservoirs corresponding to the high condition number are pruned. Simulation results indicate that PDSM-ESN has better prediction performance and run-time complexity compared with the traditional ESN models.
查看更多>>摘要：The evaluation of unmeasurable quantities is an issue faced in several fields. One example is in the production of pharmaceuticals, in which, typically, isomer properties related to different effects are found: one enantiomer may kill while the other may cure. Thus, strict quality control is necessary in this field, which requires accurate, reliable and frequent measurements of the system. However, the measurement of the main quality properties associated with the production of some pharmaceuticals has a low frequency. Consequently, in addition to safety-related issues, financial losses are also related to these low frequency of measurements, as the product can easily run towards an out of spec production. In this scenario, the present work proposes a novel Deep Artificial Intelligence structure, which has an intrinsic (Nonlinear Output Error) NOE structure, associated with a (Nonlinear AutoRegressive with Exogenous input) NARX predictor, to be used as an online soft sensor in order to provide information about the main properties of a Simulated Moving Bed chromatographic unit, commonly used in the production of pharmaceuticals, in order to mitigate the low frequency of measurement associated with this unit. The proposed structure is here called Improved Quasi-Virtual Analyzer. The model can adapt itself as the process evolves, having the possibility of online learning through measurements obtained in the laboratory periodically. The proposed structure was tested in a software-in-the-loop scenario and compared with a more traditional alternative. These tests showed a robust capacity of the Improved Quasi-Virtual Analyzer to provide reliable predictions in real-time, as well as to outperform the traditional artificial network structure.
查看更多>>摘要：Phase space reconstruction (PSR) is an effective method for chaotic system modeling, which can reveal the implicit evolution information in a complex system. However, the reconstructed time series tend to have a high dimension and contain some redundant information. It is difficult for a traditional simple model to directly forecast the reconstructed time series. In this paper, we propose a hybrid model of stacked autoencoder (SAE) and modified particle swarm optimization (MPSO) for multivariate chaotic time series forecasting. We utilize SAE to extract the reconstructed time series and adopt feedforward neural network (FNN) to forecast time series. In the proposed hybrid model, the SAE is followed by FNN, and we make the MPSO to train the output weights of the model, which is a large-scale optimization problem. To enhance the generalization ability and prevent over-fitting, we add a regularization item to the objective function when MPSO trains the weights of the model. Experimental results show that MPSO algorithm has advantages in the exploration and exploitation in large-scale optimization problems. Then, experiments on Lorenz time series and two real-world time series datasets verify the effectiveness of the hybrid model in multivariate chaotic time series forecasting.
NSTL
Elsevier