A Survey of Classification Methods in Data Streams

By Mohamed Medhat Gaber, Arkady B. Zaslavsky, Shonali Priyadarsini Krishnaswamy. 2007

111 citations recorded by the source corpus.

With the advance in both hardware and software technologies, automated data generation and storage has become faster than ever. Such data is referred to as data streams. Streaming data is ubiquitous today and it is often a challenging task to store, analyze and visualize such rapid large volumes of data. Most conventional data mining techniques have to be adapted to run in a streaming environment, because of the underlying resource constraints in terms of memory and running time. Furthermore, the data stream may often show concept drift, because of which adaptation of conventional algorithms becomes more challenging. One such important conventional data mining problem is that of classification. In the classification problem, we attempt to model the class variable on the basis of one or more feature variables. While this problem has been extensively studied from a conventional mining perspective, it is a much more challenging problem in the data stream domain. In this chapter, we will re-visit the problem of classification from the data stream perspective. The techniques for this problem need to be thoroughly re-designed to address the issue of resource constraints and concept drift. This chapter reviews the state-of-the-art techniques in the literature along with their corresponding advantages and disadvantages.

Subfields: Artificial Intelligence, Signal Processing

How researchers see this paper

The researchers whose trusted networks rank it highest, strongest advocate first. Every researcher gets a different order.

In Artificial Intelligence

In Signal Processing

Citations in the corpus

Memory Efficient Experience Replay for Streaming Learning (2018, background): “However, none of these methods learn shared representations or constrain resource usage [25].”
Lifelong Machine Learning with Deep Streaming Linear Discriminant Analysis (2019, background): “However, both Hoeffding Decision Trees and ensemble methods are slow to train [19], making them ill-suited choices for many embedded applications operating in real-time.”
On expressiveness and uncertainty awareness in rule-based classification for data streams (2017, methodology): “This is diﬀerent in DSM as the classiﬁcation model needs to be induced in a linear or sublinear time complexity [4].”
Stream-51: Streaming Classification and Novelty Detection from Videos (2020, background): “Online learning in a single pass with severe memory and compute constraints, also called streaming learning [2, 5, 10, 13, 14, 20, 35], more closely matches how embedded agents must learn and make inferences in real-time.”
Introduction to stream: An Extensible Framework for Data Stream Clustering Research with R (2017, methodology): “For a detailed discussion of these and other methods we refer the reader to the survey by Gaber, Zaslavsky, and Krishnaswamy (2007).”
The ClusTree: indexing micro-clusters for anytime stream mining (2011, background): “Approaches for frequent itemset mining [10,11,21] and classification [1,15] have been proposed.”
Online Machine Learning in Big Data Streams (2018, background): “Our list of online classiﬁcation methods is not comprehensive; for an extended list, see the recent survey [91].”
Ensemble learning for data stream analysis: A survey (2017, background): “45, 65, 116], and some have been written several years ago [109, 111, 60].”
Combination of information entropy and ensemble classification for detecting concept drift in data stream (2018, background): “Data Stream Mining have three important challenges [1-3], Speed, Size and Variability.”
SVM Parameter Optimization using Grid Search and Genetic Algorithm to Improve Classification Performance (2016, background): “Classification is a supervised learning technique which learns a function from training data set that consists of input features/attributes and categorical output [1].”