Abstract Imbalanced learning constitutes one of the most formidable challenges within data mining and machine learning. Despite continuous research advancement over past decades, from with an imbalanced class distribution remains a compelling area. distributions commonly constrain practical utility even deep models in tangible applications. Numerous recent studies have made substantial progress field learning, deepening our understanding its nature while concurrently unearthing new...

Feature selection is a highly regarded research area in the field of data mining, as it significantly enhances efficiency and performance high-dimensional analysis by eliminating redundant irrelevant features. Despite ease acquisition, labeling remains laborious expensive task. To leverage abundance unlabeled data, researchers have proposed various feature methods that operate with limited labels, including semi-supervised unsupervised selection. However, comprehensive review encompassing...

The class imbalance problem has become a leading challenge. Although conventional learning methods are proposed to tackle this problem, they have some limitations: 1) undersampling suffer from losing important information and 2) cost-sensitive sensitive outliers noise. To address these issues, we propose hybrid optimal ensemble classifier framework that combines density-based cost-effective through exploring state-of-the-art solutions using multi-objective optimization algorithm....

Broad learning system (BLS) is a novel and efficient model, which facilitates representation classification by concatenating feature nodes enhancement nodes. In spite of the properties, BLS still suboptimal when facing with imbalance problem. Besides, outliers noises in imbalanced data remain challenge for BLS. To address above issues, this paper we first propose weighted BLS, assigns weight to each training sample, adopt general weighting scheme, augments samples from minority class....

With the vigorous development of Industry 4.0, industrial Big Data has turned into core element Industrial Internet Things. As one most fundamental and indispensable components in cyber-physical systems (CPS), intelligent anomaly detection is still an essential challenging issue. However, with network, there may exist unknown types attacks, which are difficult to collect. Facing one-class intrusion scenario that collected training data only includes normal state, broad learning system...

External intrusion incidents pose a severe threat to pipeline security in energy transportation. In response this, distributed optical fiber sensing technology has been widely studied the field of safety monitoring recent years. However, diversity environment along long-distance makes vibration signal complex and changeable, which significantly limits recognition accuracy practical applications, resulting numerous false positives. To address above issues, we transform detection into...

With effective performance and fast training speed, broad learning system (BLS) has been widely developed in recent years, which provides a new way for network training. However, the randomly generated feature nodes enhancement BLS may have redundant inefficient features, will affect subsequent classification performance. In response to above issues, we propose series of self-encoding networks based on from perspective unsupervised extraction. These include single hidden layer autoencoder...

Anomaly detection stands as a crucial aspect of time series analysis, aiming to identify abnormal events in samples. The central challenge this task lies effectively learning the representations normal and patterns label-lacking scenario. Previous research mostly relied on reconstruction-based approaches, restricting representational abilities models. In addition, most current deep learning-based methods are not lightweight enough, which prompts us design more efficient framework for anomaly...

Broad learning system (BLS) are widely used due to their speed and versatility. Despite efficiency, minority class samples' accuracy is sometimes overlooked when dealing with severely imbalanced rate data. Traditional weighted BLS only considers the number of samples, such a fixed weighting leads poor classification performance. In addition, original does not take into account dispersion after its random data mapping. To solve aforementioned concerns, this study presents minimum variance...

Time-series anomaly detection has gained considerable prominence in numerous practical applications across various domains. Nonetheless, the scarcity of labels leads to neglect anomalous patterns data, as well inherent complexities and variances definitions temporal anomalies, pose significant challenges for insufficient recognition patterns. In addition, real-time poses high demands on low computational cost model robustness, presenting substantial obstacles unsupervised time-series...

The class imbalance problem has posed a leading challenge in real-world applications. Traditional methods focus on either the data level or algorithm to solve binary classification imbalanced data, and seldom consider searching an effective transformation for classification. Besides, undersampling process adopted them is always subjective unilateral. To address above issues, we first propose hybrid classifier ensemble (HCE) framework conduct classification, which mainly includes metric-based...

Ensemble clustering has an advantage in producing a more promising and robust result by combining multiple partitions strategically. The quality of both base co-association matrix plays essential role improving the consensus partition. However, current ensemble methods have several limitations: 1) noise high-dimensional feature space is ignored; 2) independent partition generation process does not pay attention to ambiguous samples; 3) weights commonly lack theoretical optimization. In order...

Broad learning system (BLS) is a simple yet efficient algorithm that only needs to train three-layer feedforward neural network. Although various BLS variants have been designed for supervised learning, none used unsupervised learning. This paper proposes BLS-AE, novel data clustering scheme seamlessly combines and auto-encoder. Then, graph regularization introduced into BLS-AE increase the capability of intrinsic structures in adaptation simultaneously, which termed BLSg-AE. Moreover,...

Efficiently capturing the complex spatiotemporal representations from large-scale traffic data with uneven quality remains to be a challenging task. In considering of dilemma, this work employs advanced contrastive learning and proposes novel Spatial-Temporal Synchronous Contextual Contrastive Learning (STS-CCL) model. First, we elaborate basic strong augmentation methods for graph data. Second, introduce Module (STS-CM) simultaneously capture decent spatial-temporal dependencies realize...

The application of artificial intelligence technology has greatly enhanced and fortified the safety energy pipelines, particularly in safeguarding against external threats. predominant methods involve integration intelligent sensors to detect vibration, enabling identification event types locations, thereby replacing manual detection methods. However, practical implementation exposed a limitation current - their constrained ability accurately discern spatial dimensions signals, which...

Time series anomaly detection is the process of identifying anomalies within time data. The primary challenge this task lies in necessity for model to comprehend characteristics time-independent and abnormal data patterns. In study, a novel algorithm called adaptive memory broad learning system (AdaMemBLS) proposed detection. This leverages rapid inference capabilities bank's capacity differentiate between normal Furthermore, an incremental based on multiple augmentation techniques...