Detecting concept drift using HEDDM in data stream

Long short-term memory self-adapting online random forests for evolving data stream regression

Anomaly detection in streaming data is an important task for many real-world applications, such as network security, fraud detection, and system monitoring. However, streaming data often exhibit concept drift, which means that the data distribution changes over time. This poses a significant challenge for many anomaly detection algorithms, as they need to adapt to the evolving data to maintain high detection accuracy. Existing streaming anomaly detection algorithms lack a unified evaluation framework that validly assesses their performance and robustness under different types of concept drifts and anomalies. In this paper, we conduct a systematic technical review of the state-of-the-art methods for anomaly detection in streaming data. We propose a new data generator, called SCAR (Streaming data generator with Customizable Anomalies and concept dRifts), that can synthesize streaming data based on synthetic and real-world datasets from different domains. Furthermore, we adapt four static anomaly detection models to the streaming setting using a generic reconstruction strategy as baselines, and then compare them systematically with 9 existing streaming anomaly detection algorithms on 76 synthesized datasets that have various types of anomalies and concept drifts. The challenges and future research directions for anomaly detection in streaming data are also presented.

Data stream mining has been receiving increased attention due to its presence in a wide range of applications, such as sensor networks, banking, and telecommunication. One of the most important challenges in learning from data streams is reacting to concept drift, i.e., unforeseen changes of the stream's underlying data distribution. Several classification algorithms that cope with concept drift have been put forward, however, most of them specialize in one type of change. In this paper, we propose a new data stream classifier, called the Accuracy Updated Ensemble (AUE2), which aims at reacting equally well to different types of drift. AUE2 combines accuracy-based weighting mechanisms known from block-based ensembles with the incremental nature of Hoeffding Trees. The proposed algorithm is experimentally compared with 11 state-of-the-art stream methods, including single classifiers, block-based and online ensembles, and hybrid approaches in different drift scenarios. Out of all the compared algorithms, AUE2 provided best average classification accuracy while proving to be less memory consuming than other ensemble approaches. Experimental results show that AUE2 can be considered suitable for scenarios, involving many types of drift as well as static environments.

A comparative study on concept drift detectors

New challenges have emerged in data mining as the traditional techniques have floundered with real-time data streams. The traditional technique needs refurbishing so as to acclimatize with concept drifting data streams. Thus dealing with the concept changes is the most imperative task of stream data mining. Ensemble classifiers have the ability to automatically adapt with the incoming drifts and, therefore, it is the most interesting research area in data stream mining. Bagging, Boosting and Random forest generation are the common ensemble techniques and are the most popular machine learning approaches in the current scenario for static data (Gomes HM, Bifet A, Read J, Barddal JP, Enembreck F, Pfharinger B, Abdessalem T (2017) Adaptive random forests for evolving data stream classification. Mach Learn 106(9–10):469–1495, [1]). A large number of base classifiers in an ensemble can cause computational overhead. Data mining classifiers for real-time data streams, therefore, need to be updated constantly and retrained with the labeled instances of the newly arrived novel classes in data streams and to cope with concept drift; otherwise, the mining models will become less and less accurate as time passes by. However, for data streams, adaptive random forest algorithms have been widely used for ensemble generation due to its competence to handle different types of drifts. This paper proposes a modified adaptive random forest with meta level learner algorithm and concept adaptive very fast decision tree to overcome the concept drift problem in real-time data streams. The proposed algorithm is experimentally compared with state-of-the-art adaptive random forest algorithm on several real synthetic datasets. Results indicate its efficiency in terms of accuracy and processing time.

We present a novel method for concept drift detection, based on: 1) the development and continuous updating of online sequential extreme learning machines (OS-ELMs) and 2) the quantification of how much the updated models are modified by the newly collected data. The proposed method is verified on two synthetic case studies regarding different types of concept drift and is applied to two public real-world data sets and a real problem of predicting energy production from a wind plant. The results show the superiority of the proposed method with respect to alternative state-of-the-art concept drift detection methods. Furthermore, updating the prediction model when the concept drift has been detected is shown to allow improving the overall accuracy of the energy prediction model and, at the same time, minimizing the number of model updatings.

Machine learning algorithms have been widely used in IoT application systems for predicting labels of input data streams from IoT sensors. Since data streams are not always stationary in real-world scenarios, underlying data distribution changes may cause a deterioration of the prediction performance that is known as concept drift. Existing concept drift detection methods often assume that complete and true labels for detecting the prediction errors are always available immediately after the prediction. However, such an assumption is not realistic in real IoT application systems. This paper experimentally investigates the robustness of six representative concept drift detectors against unreliable data streams containing the data with error labels and the data with no labels. Robustness is evaluated in terms of the average detection delay and precision by increasing the ratios of error-labeled and missing-label data in a synthetic data stream. Our experimental results show that Cumulative Sum (CUSUM), Page-Hinkley (PH), and Drift Detection Method (DDM) achieve relatively stable performances when the ratio of error-labeled data is less than 40%. With respect to the drift detection efficiency, CUSUM can be regarded as the most robust detector among other detectors tested in our experiments.

Neural network models are widely adopted in real-world applications for processing streaming data. However, these applications often face challenges in terms of accuracy degradation, caused by changes in the data distribution of the stream data compared to the training data. Two underlying reasons contribute to these changes. The first, known as the concept drift problem, occurs when there is a change in the correlation between the input data and the prediction output, making the models trained on the training data no longer suitable for the new data. The second reason, known as the novelty problem, arises when real-world data contains unexpected data categories that were not present in the training data, resulting in incorrect predictions. The research community has divided into different groups and each developed various methods to detect either concept drift or novelty distribution changes. However, these methods only address one aspect of the problem and are unable to distinguish between them. This leads to an inappropriate allocation of model maintenance resources, including the high cost of model retraining and the acquisition of true label data. In this study, we aim to address this gap by proposing a novel concept drift detection method that is capable of distinguishing between known labeled concept drift and novelty. Our method is also more efficient than existing drift detection methods, making it suitable for applications on neural networks.

Selective ensemble-based online adaptive deep neural networks for streaming data with concept drift

Concept drifts usually originate from many causes instead of only one, which result in two types of concept drifts: abrupt drifts and gradual drifts. From the point of view of speed, concept drifts pose strong challenges for data stream mining. In this paper, we propose a selective detector ensemble to detect both abrupt and gradual drifts. We first present our detector ensemble construction method, and then introduce how to use this ensemble to detect concept drifts with the proposed early-findearly-report rule.To evaluate the performance of our method, we compare it with four drift detection methods on eight publicly available data sets containing various concept drifts. The experimental results show that compared with those benchmarks, our ensemble method can effectively improve the recall and false negative rate without significantly increasing the false positive rate, and has stronger generalization ability than those single-change-indicator-based methods.

The advances in computing software, hardware, connected devices and wireless communication infrastructure in recent years have led to the desire to work with streaming data sources. Yet the number of techniques, approaches and algorithms which can work with data from a streaming source is still very limited, compared with batched data. Although data mining techniques have been a well-studied topic of knowledge discovery for decades, many unique properties as well as challenges in learning from a data stream have not been considered properly due to the actual presence of and the real needs to mine information from streaming data sources. This thesis aims to contribute to the knowledge by developing a rule-based algorithm to specifically learn classification rules from data streams, with the learned rules are expressive so that a human user can easily interpret the concept and rationale behind the predictions of the created model. There are two main structures to represent a classification model; the ‘tree-based’ structure and the ‘rule-based’ structure. Even though both forms of representation are popular and well-known in traditional data mining, they are different when it comes to interpretability and quality of models in certain circumstances. The first part of this thesis analyses background work and relevant topics in learning classification rules from data streams. This study provides information about the essential requirements to produce high quality classification rules from data streams and how many systems, algorithms and techniques related to learn the classification of a static dataset are not applicable in a streaming environment. The second part of the thesis investigates at a new technique to improve the efficiency and accuracy in learning heuristics from numeric features from a streaming data source. The computational cost is one of the important factors to be considered for an effective and practical learning algorithm/system because of the needs to learn from continuous arrivals of data examples sequentially and discard the seen data examples. If the computing cost is too expensive, then one may not be able to keep pace with the arrival of high velocity and possibly unbound data streams. The proposed technique was first discussed in the context of the use of Gaussian distribution as heuristics for building rule terms on numeric features. Secondly, empirical evaluation shows the successful integration of the proposed technique into an existing rule-based algorithm for the data stream, eRules. Continuing on the topic of a rule-based algorithm for classification data streams, the use of Hoeffding’s Inequality addresses another problem in learning from a data stream, namely how much data should be seen from a data stream before starting learning and how to keep the model updated over time. By incorporating the theory from Hoeffding’s Inequality, this study presents the Hoeffding Rules algorithm, which can induce modular rules directly from a streaming data source with dynamic window sizes throughout the learning period to ensure the efficiency and robustness towards the concept drifts. Concept drift is another unique challenge in mining data streams which the underlying concept of the data can change either gradually or abruptly over time and the learner should adapt to these changes as quickly as possible. This research focuses on the development of a rule-based algorithm, Hoeffding Rules, for data stream which considers streaming environments as primary data sources and addresses several unique challenges in learning rules from data streams such as concept drifts and computational efficiency. This knowledge facilitates the need and the importance of an interpretable machine learning model; applying new studies to improve the ability to mine useful insights from potentially high velocity, high volume and unbounded data streams. More broadly, this research complements the study in learning classification rules from data streams to address some of the unique challenges in data streams compared with conventional batch data, with the knowledge necessary to systematically and effectively learn expressive and modular classification rules from data streams.

One of the most important challenges for machine learning community is to develop efficient classifiers which are able to cope with data streams, especially with the presence of the so-called concept drift. This phenomenon is responsible for the change of classification task characteristics, and poses a challenge for the learning model to adapt itself to the current state of the environment. So there is a strong belief that one-class classification is a promising research direction for data stream analysis--it can be used for binary classification without an access to counterexamples, decomposing a multi-class data stream, outlier detection or novel class recognition. This paper reports a novel modification of weighted one-class support vector machine, adapted to the non-stationary streaming data analysis. Our proposition can deal with the gradual concept drift, as the introduced one-class classifier model can adapt its decision boundary to new, incoming data and additionally employs a forgetting mechanism which boosts the ability of the classifier to follow the model changes. In this work, we propose several different strategies for incremental learning and forgetting, and additionally we evaluate them on the basis of several real data streams. Obtained results confirmed the usability of proposed classifier to the problem of data stream classification with the presence of concept drift. Additionally, implemented forgetting mechanism assures the limited memory consumption, because only quite new and valuable examples should be memorized.

Concept drift referring to the changes of data distributions has been one critical challenge typically associated with mining data streams. Current drift detection and adaptation methods focus on how to immediately detect the distribution changes once the concept drift occurs and swiftly update the model to be applicable to the newly arrived data instances. Most of those methods assume the data does not have noise or the noise is too weak to affect the modeling procedure. However, realworld data are normally contaminated, and denoise techniques are highly preferred as a necessary preprocess. This issue is more complex for a data stream with concept drift because the noise is very likely to be confused with drift. Motivated by that, this paper proposes a Noise-tolerant Fuzzy c-means based drift Adaptation method (NFA) which can adapt to the changing distributions and is suitable for noisy data streams. The concept drift problem is solved by using a fuzzy c-means based regression model to continuously include the most relevant data instances to the latest pattern in the training set. In addition, a denoise technique is designed in NFA to remove noise, and the ability of incremental updating enables it to be embedded in the incremental drift adaptation process, and therefore NFA can solve concept drift and noise problems at the same time. Experimental evaluation results also show good performance of our method on handling data streams with concept drift and noise.

A large-scale comparison of concept drift detectors

The last decade has seen a surge of interest in adaptive learning algorithms for data stream classification, with applications ranging from predicting ozone level peaks, learning stock market indicators, to detecting computer security violations. In addition, a number of methods have been developed to detect concept drifts in these streams. Consider a scenario where we have a number of classifiers with diverse learning styles and different drift detectors. Intuitively, the current 'best' (classifier, detector) pair is application dependent and may change as a result of the stream evolution. Our research builds on this observation. We introduce the $\mbox{Tornado}$ framework that implements a reservoir of diverse classifiers, together with a variety of drift detection algorithms. In our framework, all (classifier, detector) pairs proceed, in parallel, to construct models against the evolving data streams. At any point in time, we select the pair which currently yields the best performance. We further incorporate two novel stacking-based drift detection methods, namely the $\mbox{FHDDMS}$ and $\mbox{FHDDMS}_{add}$ approaches. The experimental evaluation confirms that the current 'best' (classifier, detector) pair is not only heavily dependent on the characteristics of the stream, but also that this selection evolves as the stream flows. Further, our $\mbox{FHDDMS}$ variants detect concept drifts accurately in a timely fashion while outperforming the state-of-the-art.