An Approach For Concept Drift Detection in a Graph Stream Using Discriminative Subgraphs

Abstract

The emergence of mining complex networks like social media, sensor networks, and the world-wide-web has attracted considerable research interest. In a streaming scenario, the concept to be learned can change over time. However, while there has been some research done for detecting concept drift in traditional data streams, little work has been done on addressing concept drift in data represented as a graph . We propose a novel unsupervised concept-drift detection method on graph streams called Discriminative Subgraph-based Drift Detector (DSDD). The methodology starts by discovering discriminative subgraphs for each graph in the stream. We then compute the entropy of the window based on the distribution of discriminative subgraphs with respect to the graphs and then use the direct density-ratio estimation approach for detecting concept drift in the series of entropy values obtained by moving one step forward in the sliding window. The effectiveness of the proposed method is demonstrated through experiments using artificial and real-world datasets and its performance is evaluated by comparing against related baseline methods. Similarly, the usefulness of the proposed concept drift detection approach is studied by incorporating it in a popular graph stream classification algorithm and studying the impact of drift detection in classification accuracy.

Multi-domain and Explainable Prediction of Changes in Web Vocabularies

“…In the future, we plan to incorporate and compare directly our results with other approaches [11,25,28]; this comparison was impossible since these approaches do not focus on version chains as we do. Specifically, we will apply our framework to ontologies from the biomedical domain [27] and use the same dataset as [26] to compare performance more directly.…”

“…However, their focus is not on the change prediction but rather on the update which follows. Robust learning algorithms in ontology streams with semantic drift have also been investigated [5,25]; however, graph streams demand a much higher update frequency than the classic ontology versions we study here.…”

Online embedding and clustering of evolving data streams

“…There exist several concept drift detection algorithms in the literature. Among the most recent drift detection algorithms are one class drift detector [18], discriminative drift detector (D3) [19], accurate concept drift detection method [20] and discriminative subgraph-based drift detector [21]. Gama et al [22] have a comprehensive survey on concept drift detection.…”

Human-in-the-loop handling of knowledge drift

“…Drift detection has been studied in the context of graph classification. In this setting, the input data streams encodes a sequence of knowledge graphs (Paudel and Eberle 2020;Yao and Holder 2016;Zambon et al 2018) or an ontology stream (Chen et al 2017), and drift affects the distribution of said graphs. In contrast, in our target applications the machine receives a sequence of examples, consisting of a set of subsymbolic observations and of concept annotations, while the knowledge graph controlling the relationship between concepts is completely unobserved.…”