A feedforward neural network ensemble trained through metaheuristic algorithms has been proposed by researchers to produce a group of optimal neural networks. This method, however, has proven to be very time-consuming during the optimization process. To overcome this limitation, we propose a metaheuristic-based learning algorithm for building an ensemble system, resulting in shorter training time. In our proposed method, a master-slave based metaheuristic algorithm is employed in the optimization process to produce a group of heterogeneous feedforward neural networks, in which the global search operations are executed on the master, and the tasks of objective evaluation are distributed to the slaves (workers). To reduce evaluation costs, the entire training dataset is randomly divided equally into several disjoint subsets. Each subset is randomly paired with another subset of the remainder and distributed to a worker for the objective evaluation. Following the optimization process, representative candidate solutions (individuals) from the entire population are selected to perform as the base components of the ensemble system. The performance of the proposed method has been compared with those of other state-of-the-art techniques in over 31 benchmark regression datasets taken from public repositories. The experimental results show that the proposed method not only reduces the computational time but also achieves significantly better prediction accuracy. Moreover, the proposed method achieved promising results in the application of a subset of the million song dataset, which identifies the release year of a song and predicts the buzz on Twitter.

Owing to the exponential proliferation of internet services and the sophistication of intrusions, traditional intrusion detection algorithms are unable to handle complex invasions due to their limited representation capabilities and the unbalanced nature of Internet of Things (IoT)-related data in terms of both telemetry and network traffic. Drawing inspiration from deep learning achievements in feature extraction and representation learning, in this study, we propose an accurate hybrid ensemble deep learning framework (HEDLF) to protect against obfuscated cyber-attacks on IoT networks . To address complex features and alleviate the imbalance problem, the proposed HEDLF includes three key components: (1) a hierarchical feature representation technique based on deep learning, which aims to extract specific information by supervising the loss of gradient information; (2) a balanced rotated feature extractor that simultaneously encourages the individual accuracy and diversity of the ensemble classifier; and (3) a meta-classifier acting as an aggregation method, which leverages a semisparse group regularizer to analyze the base classifiers' outputs. Additionally, these improvements take class imbalance into account. The experimental results show that when compared against state-of-the-art techniques in terms of accuracy, precision, recall, and F1-score, the proposed HEDLF can achieve promising results on both telemetry and network traffic data.

Multi-label learning with emerging new labels is a practical problem that occurs in data streams and has become an important new research issue in the area of machine learning. However, existing models for dealing with this problem require high learning computational times, and there still exists a lack of research. Based on these issues, this paper presents an incremental kernel extreme learning machine for multi-label learning with emerging new labels, consisting of two parts: a novelty detector; and a multi-label classifier. The detector with free-user-setting threshold parameters was developed to identify instances with new labels. A new incremental multi-label classifier and its improved version were developed to predict a label set for each instance, which can add output units incrementally and update themselves in unlabeled instances. Comprehensive evaluations of the proposed method were carried out on the problems of multi-label classification with emerging new labels compared to comparative algorithms, which revealed the promising performance of the proposed method.

MOOCs are online learning environments which many students use, but the success rate of online learning is low. Machine learning can be used to predict learning success based on how people learn in MOOCs. Predicting the learning performance can promote learning through various methods, such as identifying low-performance students or by grouping students together. Recent machine learning has enabled the development of predictive models, and the ensemble method can assist in reducing the variance and bias errors associated with single-machine learning. This study uses a two-phase classification model with an ensemble technique to predict the learners’ grades. In the first phase, binary classification is used, and the non-majority class is then sent to the second phase, which is multi-class classification. The new features are computed based on the distance from the class’s center. The distance between the data and the center of an overlapping cluster is calculated using silhouette score-based feature selection. Lastly, Bayesian optimization boosts the performance by fine tuning the optimal parameter set. Using data from the HMPC- and the CNPC datasets, the experiment results demonstrate that the proposed design, the two-phase ensemble-based method, outperforms the state-of-the-art machine learning algorithms.

In light of the flourishing proliferation of internet services, the popularity of the Internet of Things (IoT) has swiftly grown in the medical and healthcare fields, and this has been accompanied by a simultaneous escalation in the sophistication of intrusion attacks. Drawing inspiration from the accomplishments of deep learning in cyber threat detection, we propose a multigrained scanning-based deep stacking network (MGDSN) to defend against sophisticated cyberattacks on Internet of Medical Things (IoMT) networks. To address the obscured characteristics of intricate cyberattacks, the MGDSN incorporates four components. First, the feature augmentation process leverages an improved multigrained scanning technique to enhance discriminative information. Second, a deep stacking network (DSN) with a weighting mechanism is employed to generate a set of predictive results for making the final decision. Third, a meta-classifier is introduced to scrutinize the influence of the predictive results when producing the final decision and exploiting a set of meaningfully extracted features. Finally, a loss function is properly designed to take both the predictive losses of the DSN modules and the final predictive loss into account. The outstanding performance achieved by the MGDSN is confirmed through comprehensive evaluations comparing it with the state-of-the-art techniques, encompassing metrics such as the accuracy, precision, recall, F1 score, Cohen’s kappa coefficient, Matthews correlation coefficient, and area under the curve achieved on the IoMT datasets. The MGDSN exhibits a notable improvement ranging from approximately 0.12%-329.21%.