Unplanned train service disruptions are a major concern for cities with high public transport ridership. Train service disruptions causes rapid degradation of level of service provided by transit systems. Understanding the travel choices made by transit passengers during service disruptions is a crucial step in managing such disruptions. Using two revealed preference datasets from Singapore that include 1) data from a questionnaire survey that asked participants about their travel plan choices during a major service disruption and 2) transit smart card data that provided the transit routes taken by commuters during a major service disruption, this study models the decision making process of commuters faced with unplanned train service disruption. We propose a three stage sequential decision making framework, that include three separate Multinomial logit (MNL) models to predict the decision to travel, the mode choice and the transit route choice. The data and the estimated models show that the commuters have high preference to go on with their journey plan using alternative public transit routes, in particular by using a bridging bus service to substitute the disrupted segment of the train network. The estimated models can be used to design effective mitigation strategies for unplanned train service disruptions.