Online Clustering for Novelty Detection and Concept Drift in Data Streams

Abstract

Data streams are related to large amounts of data that can continuously arrive with a probability distribution that may change over time. Depending on the changes in the data distribution, different phenomena can occur, like new classes can appear or concept drift can occur in existing classes. Machine Learning algorithms have been often used to model this data. New classes are patterns that were not seen during the training of the current classification model, but appear after some time. Concept drift occurs when the concepts associated with a dataset change as new data arrive. This paper proposes a new algorithm based on k NN that uses micro-clusters as prototypes and incrementally updates the micro-clusters or creates new micro-clusters when novelties are detected. In the online phase, each instance close to a micro-cluster is considered an extension of the micro-cluster, being used to adapt the model to concept drift. The proposed algorithm is experimentally compared with a stateof-the-art classifier from the data stream literature and one baseline. According to the experimental results, the proposed algorithm increases the predictive performance over time by incrementally learning changes in the data distribution.

Feature-Based Multi-Class Classification and Novelty Detection for Fault Diagnosis of Industrial Machinery

“…In these cases, the training procedure is the same as the multi-class classification. In other cases, unsupervised learning-based models, such as clustering models, are used when novelties need to be recognized in streaming data [22,33]. Clustering approaches can be based on distance measures or density measures.…”

“…Another approach to novelty detection sees novelty as a concept, an abstraction of cohesive and representative examples, that introduces characteristics different from known concepts [21]. In this sense, several clustering algorithms for novelty detection in data streams have been proposed, such as OnLine Novelty Detection and Drift Detection Algorithm (OLINDDA) [21], Higia [22], and MINAS (MultIclass learNing Algorithm for data Streams) [12]. The idea behind these algorithms is to learn a decision model based on labeled data during the offline training phase; then, during the online phase, novelty patterns are recognized as unknown sets, or micro-clusters, made of examples that the model does not explain.…”

Unsupervised Learning Approaches for Non-Stationary Data Streams

“…• Garcia, K. D., Poel, M., Kok, J. N., & de Carvalho, A. C. (2019, September). Online Clustering for Novelty Detection and Concept Drift in Data Streams.…”

Unsupervised learning approaches for non-stationary data streams

“…Chapter 4, Online Clustering for Novelty Detection and Concept Drift in Data Streams [48], presents an unsupervised approach for concept changes in data streams. This paper was published in the proceedings of the EPIA 2019 conference.…”