In most of the analyses using the cosmic muon data from the INO-ICAL prototype stack, only single muon events are considered. Multi-muon events appear to be noisy events to the algorithm and thus get rejected, reducing the physics potential of the detector. To address this issue, we have developed an ML algorithm to predict muon multiplicity in cosmic muon event registered in the INO-ICAL detector at TIFR. In this work, we present the performance of this algorithm in terms of its efficiency to correctly tag multi-muon events and also to predict the multiplicity