$k$-means clustering for predicting the period of fast radio bursts

作者：Zhengxiang Li · 发表于：Research in Astronomy and Astrophysics · 年份：2026 · DOI：10.1088/1674-4527/ae84f9

Fast radio bursts (FRBs) are one of the most enigmatic phenomena in modern astronomy. In observations, these transient radio emissions can be categorized into two distinct classes: repeaters and non-repeaters. For repeaters, periodic activities play an important role in understanding their progenitors and radiation mechanisms. To date, only two FRBs, FRB 20121102 and FRB 20180916B, have been confirmed to exhibit cyclical activity, with periods of approximately 160 days and 16 days, respectively. Recently, FRB 20240209A has been identified as a potential candidate with periodicity of approximately four months. In this study, we propose to use the $k$-means clustering algorithm to derive possible periods of active repeaters from bursting time observations. Moreover, the method successfully re-detects the known period of FRB 20180916B, as well as the previously reported candidate periods of FRB 20121102A and FRB 20240209A. These findings are in excellent agreement with previous results and a strong indication for the potential of machine learning approaches in uncovering hidden patterns in FRB data, which may otherwise remain obscured by observational noise and complexity. As more observational data become available and machine learning methodologies continue to evolve, such computational tools are expected to play an increasingly crucial role in advancing our understanding of FRBs, potentially revealing new insights into their physical origins, environmental contexts, and under...