Gamma-ray burst (GRB) light curves contain a remarkable level of temporal complexity, reflecting both the physics of the relativistic outflow and potential gravitational distortions caused by intervening compact objects. Gravitational microlensing, in particular, can imprint subtle yet measurable modifications on the temporal profile of a GRB. A central challenge, however, lies in reliably detecting these signatures within large GRB data sets, where the differences between lensed and non-lensed events may be difficult to identify by inspection alone.

In this study, we demonstrate that the temporal features inherently present in GRB light curves carry sufficient information to distinguish microlensed bursts from their unlensed counterparts. To explore this, we generate an extensive suite of simulated GRB light curves representing both lensed and non-lensed classes. Using the Cesium package and other feature extraction tools, we extract a comprehensive set of features that characterize key aspects of the temporal behavior, including variability patterns, statistical moments, and other descriptors of morphological evolution. These features serve as input to a machine learning (ML) pipeline designed to identify lensing-induced signatures.

We evaluate several ML algorithms on this feature space. Our results demonstrate that light curves themselves contain rich and discriminative information,and that feature-based ML approaches are capable of exploiting this information effectively. Our work highlights the potential of such methods for future observational applications, including automated searches for GRB lensing.