Random forests is currently one of the most used machine learning algorithms in the non-streaming (batch) setting. This preference is attributable to its high learning performance and low demands with respect to input preparation and hyper-parameter tuning. However, in the challenging context of evolving data streams, there is no random forests algorithm that can be considered state-of-the-art in comparison to bagging and boosting based algorithms.

Ensemble-based methods are among the most widely used techniques for data stream classification. Their popularity is attributable to their good performance in comparison to strong single learners while being relatively easy to deploy in real-world applications. Ensemble algorithms are especially useful for data stream learning as they can be integrated with drift detection algorithms and incorporate dynamic updates, such as selective removal or addition of classifiers. This work proposes a taxonomy for data stream ensemble learning as derived from reviewing over 60 algorithms. Important aspects such as combination, diversity, and dynamic updates, are thoroughly discussed. Additional contributions include a listing of popular open-source tools and a discussion about current data stream research challenges and how they relate to ensemble learning (big data streams, concept evolution, feature drifts, temporal dependencies, and others).

Data stream mining is a fast growing research topic due to the ubiquity of data in several real-world problems. Given their ephemeral nature, data stream sources are expected to undergo changes in data distribution, a phenomenon called concept drift. This paper focuses on one specific type of drift that has not yet been thoroughly studied, namely feature drift. Feature drift occurs whenever a subset of features becomes, or ceases to be, relevant to the learning task, thus, learners must detect and adapt to these changes accordingly. We survey existing work on feature drift adaptation in both explicit and implicit approaches. Additionally, we benchmark several algorithms and a naive proposal in synthetic and real-world datasets. The results from our experiments indicate the need for future research in this area as even naive approaches produced gains in accuracy while reducing resources usage. Finally, we state current research topics, challenges and future directions for feature drift adaptation.

Incremental learning, online learning, and data stream learning are terms commonly associated with learning algorithms that update their models given a continuous influx of data without performing multiple passes over data. Several works have been devoted to this area, either directly or indirectly as characteristics of big data processing, i.e., Velocity and Volume. Given the current industry needs, there are many challenges to be addressed before existing methods can be efficiently applied to real-world problems. In this work, we focus on elucidating the connections among the current stateof- the-art on related fields; and clarifying open challenges in both academia and industry. We treat with special care topics that were not thoroughly investigated in past position and survey papers. This work aims to evoke discussion and elucidate the current research opportunities, highlighting the relationship of different subareas and suggesting courses of action when possible.

No abstract

No abstract

Feature selection targets the identification of which features of a dataset are relevant to the learning task. It is also widely known and used to improve computation times, reduce computation requirements, and to decrease the impact of the curse of dimensionality and enhancing the generalization rates of classifiers. In data streams, classifiers shall benefit from all the items above, but more importantly, from the fact that the relevant subset of features may drift over time. In this paper, we propose a novel dynamic feature selection method for data streams called Adaptive Boosting for Feature Selection (ABFS). ABFS chains decision stumps and drift detectors, and as a result, identifies which features are relevant to the learning task as the stream progresses with reasonable success. In addition to our proposed algorithm, we bring feature selection-specific metrics from batch learning to streaming scenarios. Next, we evaluate ABFS according to these metrics in both synthetic and real-world scenarios. As a result, ABFS improves the classification rates of different types of learners and eventually enhances computational re

No abstract