Online non-parametric change-point detection for heterogeneous data streams observed over graph nodes

TL;DR

This paper introduces an online non-parametric change-point detection method for heterogeneous graph-based data streams, leveraging kernel techniques to identify distribution changes across nodes in real-time.

Contribution

It presents a novel kernel-based, non-parametric approach for detecting change-points in heterogeneous graph data streams, accounting for node similarities.

Findings

Effective detection of change-points demonstrated on synthetic data.

Successful application to real-world graph data streams.

Method outperforms existing approaches in heterogeneous settings.

Abstract

Consider a heterogeneous data stream being generated by the nodes of a graph. The data stream is in essence composed by multiple streams, possibly of different nature that depends on each node. At a given moment $\tau$, a change-point occurs for a subset of nodes $C$, signifying the change in the probability distribution of their associated streams. In this paper we propose an online non-parametric method to infer $\tau$ based on the direct estimation of the likelihood-ratio between the post-change and the pre-change distribution associated with the data stream of each node. We propose a kernel-based method, under the hypothesis that connected nodes of the graph are expected to have similar likelihood-ratio estimates when there is no change-point. We demonstrate the quality of our method on synthetic experiments and real-world applications.