Evolutionary complex network for uncovering rich structure of series

Abstract

Important structures hidden in series often reflect various real-world information. Analyzing and recognizing series is, therefore, of great practical significance. Complex networks have shown outstanding performance in mining the topological features of data, which provides rich information from high-dimensional perspective. In this work, we develop a new evolutionary complex network mapped method from series, termed weighted k-series maximum differential graph (ks-maxDG). This method facilitates the mapping of series into complex networks from multiple perspectives, providing a more comprehensive exploration of their topological properties. These dynamic network properties offer deeper insights into the evolving structure of the original series. We validate its accuracy in uncovering the topological features theoretically and empirically, showing excellent performance in chaos and noise identification as well as series classification.