The Ring Imaging CHerenkov (RICH) detector is the primary electron identification detector in the Compressed Baryonic Matter (CBM) experiment.
Conventionally a single layer perceptron based artificial neural network (ANN) is used as the electron identifier in the RICH reconstruction scheme.
The ANN takes in the Cherenkov ring parameters, track momentum, and ring-track distance as the input features and outputs the probability that the given particle is an electron or otherwise.
In this work, the conventional ANN is replaced by the gradient boosted ensemble of decision trees (XGBoost package).
Furthermore, the existing input features are modified and extended, which involve changing the 2D- ring track distance to differential 1D- distances.
Additionally, a new ring-track reference is derived by refitting the global track (which has hits in tracking station + PID detectors) in the Transition Radiation Detector (TRD) downstream to RICH and extrapolating the fitted TRD tracklets upstream to the RICH.
Furthermore, a method is developed to derive a probability that a ring stems from the photon pair production (conversion probability) in the detector material after tracking stations.
Finally, an upgraded XGBoost based electron identifier was developed using the Cherenkov ring parameters, momentum, differential (1D) ring-track distances (extrapolated from STS), differential ring-backtrack references (extrapolated from TRD) and conversion probability for the Cherenkov ring.