Practical applications involving big data, such as weather monitoring, identification of customer preferences, Internet log analysis, and sensors warnings require challenging data analysis, since these are examples of problems whose data are generated in streams and usually demand real‐time analytics. Patterns in such data stream problems may change quickly. Consequently, machine learning models that operate in this context must be updated over time. This phenomenon is called concept drift in machine learning and data mining literature. Several different directions have been pursued to learn from data stream and to deal with concept drift. However, most drift detection methods consider that an instance's class label is available right after its prediction, since these methods work by monitoring the prediction results of a base classifier or an ensemble of classifiers. Nevertheless, this constraint is unrealistic in several practical problems. To cope with this constraint, some works are focused on proposing efficient unsupervised or semi‐supervised concept drift detectors. While interesting and recent overview papers dedicated to supervised drift detectors have been published, the scenario is not the same in terms of unsupervised methods. Therefore, this work presents a comprehensive overview of approaches that tackle concept drift in classification problems in an unsupervised manner. Additional contribution includes a proposed taxonomy of state‐of‐the‐art approaches for concept drift detection based on unsupervised strategies.This article is categorized under: Technologies > Classification Technologies > Machine Learning

The use of machine learning (ML) techniques in affective computing applications focuses on improving the user experience in emotion recognition. The collection of input data (e.g., physiological signals), together with expert annotations are part of the established standard supervised learning methodology used to train human emotion recognition models. However, these models generally require large amounts of labeled data, which is expensive and impractical in the healthcare context, in which data annotation requires even more expert knowledge. To address this problem, this paper explores the use of the self-supervised learning (SSL) paradigm in the development of emotion recognition methods. This approach makes it possible to learn representations directly from unlabeled signals and subsequently use them to classify affective states. This paper presents the key concepts of emotions and how SSL methods can be applied to recognize affective states. We experimentally analyze and compare self-supervised and fully supervised training of a convolutional neural network designed to recognize emotions. The experimental results using three emotion datasets demonstrate that self-supervised representations can learn widely useful features that improve data efficiency, are widely transferable, are competitive when compared to their fully supervised counterparts, and do not require the data to be labeled for learning.

Concept drift, a phenomenon that can lead to degradation of classifier performance over time, is commonly addressed in the literature through detection and reaction strategies. However, these strategies often rely on complete classifier retraining without considering the properties of the drift, which can prove inadequate in many scenarios. Limited attention has been given to understanding the nature of drift and its characterization, which are crucial for designing effective reaction strategies. Drift descriptors provide a means to explain how new concepts replace existing ones, offering valuable insights into the nature of drift. In this context, this work examines the impact of four descriptors—severity, recurrence, frequency, and speed—on concept drift through extensive theoretical and experimental analysis. Experiments were conducted on five datasets with 32 descriptor variations, eight drift detectors, and a non-detection context, resulting in 1440 combinations. The findings reveal three key conclusions: (i) reaction strategies must be tailored to different types of drift; (ii) severity, recurrence, and frequency descriptors have the highest impact, whereas speed has minimal influence; and (iii) there is a need to incorporate mechanisms for describing concept drift into the strategies designed to address it.

Electronic messages are still considered the most significant tools in business and personal applications due to their low cost and easy access. However, e-mails have become a major problem owing to the high amount of junk mail, named spam, which fill the e-mail boxes of users. Several approaches have been proposed to detect spam, such as filters implemented in e-mail servers and user-based spam message classification mechanisms. A major problem with these approaches is spam detection in the presence of concept drift, especially as a result of changes in features over time. To overcome this problem, this work proposes a new spam detection system based on analyzing the evolution of features. The proposed method is divided into three steps: 1) spam classification model training; 2) concept drift detection; and 3) knowledge transfer learning. The first step generates classification models, as commonly conducted in machine learning. The second step introduces a new strategy to avoid concept drift: SFS (Similarity-based Features Se- lection) that analyzes the evolution of the features taking into account similarity obtained between the feature vectors extracted from training data and test data. Finally, the third step focuses on the following questions: what, how, and when to transfer acquired knowledge? The proposed method is evaluated using two public datasets. The results of the experiments show that it is possible to infer a threshold to detect changes (drift) in order to ensure that the spam classification model is updated through knowledge transfer. Moreover, our anomaly detection system is able to perform spam classification and concept drift detection as two parallel and independent tasks.

Dynamic classifier selection has traditionally focused on selecting the most accurate classifier to predict the class of a particular test pattern. In this paper we propose a new dynamic selection method to select, from a population of ensembles, the most confident ensemble of classifiers to label the test sample. Such a level of confidence is measured by calculating the ambiguity of the ensemble on each test sample. We show theoretically and experimentally that choosing the ensemble of classifiers, from a population of high accurate ensembles, with lowest ambiguity among its members leads to increase the level of confidence of classification, consequently, increasing the generalization performance. Experimental results conducted to compare the proposed method to static selection and DCS-LA, demonstrate that our method outperforms both DCS-LA and static selection strategies when a population of high accurate ensembles is available.