Hierarchical clustering for multiple nominal data streams with evolving behaviour

Over the decade, a number of attempts have been made towards data stream clustering, but most of the works fall under clustering by example approach. There are a number of applications where clustering by variable approach is required which involves clustering of multiple data streams as opposed to clustering data examples in a data stream. Furthermore, a few works have been presented for clustering multiple data streams and these are applicable to numeric data streams only. Hence, this research gap has motivated current research work. In the present work, a hierarchical clustering technique has been proposed to cluster multiple data streams where data are nominal. To address the concept changes in the data streams splitting and merging of the clusters in the hierarchical structure are performed. The decision to split or merge is based on the entropy measure, representing the cluster’s degree of disparity. The performance of the proposed technique has been analysed and compared to Agglomerative Nesting clustering technique on synthetic as well as a real-world dataset in terms of Dunn Index, Modified Hubert $$\varGamma $$ Γ statistic, Cophenetic Correlation Coefficient, and Purity. The proposed technique outperforms Agglomerative Nesting clustering technique for concept evolving data streams. Furthermore, the effect of concept evolution on clustering structure and average entropy has been visualised for detailed analysis and understanding.

Improvement of Data Stream Decision Trees

The classification of data streams has become a significant and active research area. The principal characteristics of data streams are a large amount of arrival data, the high speed and rate of its arrival, and the change of their nature and distribution over time. Hoeffding Tree is a method to, incrementally, build decision trees. Since its proposition in the literature, it has become one of the most popular tools of data stream classification. Several improvements have since emerged. Hoeffding Anytime Tree was recently introduced and is considered one of the most promising algorithms. It offers a higher accuracy compared to the Hoeffding Tree in most scenarios, at a small additional computational cost. In this work, the authors contribute by proposing three improvements to the Hoeffding Anytime Tree. The improvements are tested on known benchmark datasets. The experimental results show that two of the proposed variants make better usage of Hoeffding Anytime Tree’s properties. They learn faster while providing the same desired accuracy.