Instance-Based Classification of Streaming Data Using Emerging Patterns

Abstract

Classification of Streaming Data has been recently recognized as an important research area. It is different from conventional techniques of classification because we prefer to have a single pass over each data item. Moreover, unlike conventional classification, the true labels of the data are not obtained immediately during the training process. This paper proposes ILEP, a novel instance-based technique for classification of streaming data with a modifiable reference set based on the concept of Emerging Patterns. Emerging Patterns (EPs) have been successfully used to catch important data items for addition to the reference set, hence resulting in an increase in classification accuracy as well as restricting the size of the reference set.

Fast Unsupervised Online Drift Detection Using Incremental Kolmogorov-Smirnov Test

“…Although Kuncheva et al [20] and Amir and Toshniwal [2] introduced methods to directly tackle delayed labels, they target problems with stationary distributions, i.e., problems without concept drift. Kuncheva proposes different variations of the Nearest Neighbor Classifier (NNC) for online learning.…”

Non-stationary and unpredictable data distributions in classification and quantification

“…Such groups detect concept drifts through deviations in stored statistics. Kuncheva and Sánchez (2008), Amir and Toshniwal (2010) propose methods that directly tackle delays in the availability of true labels. However, they are limited to stationary distribution, i.e., classification problems without concept drifts.…”