In a traffic network with uncertain travel time，we postulate that the travelers are heterogeneous in terms of their reference points，which follow a continuous distribution function，and at the same time the degree of the diminishing sensitivity is dependent on the reference points．We then develop a multi-class，cumulative prospect theory-based equilibrium model by incorporating these heterogeneities．The results show the risk is treated by different users in different manners: those with higher and lower reference points prefer the routes with higher risks to get a higher perceived value，and those with medium reference points mainly focus on the routes with lower risks．In addition，a sensitivity analysis is conducted with respect to two management methods，i.e.，road expansion and risk management in improving the perceived value for each kind of users．The results show that road expansion is an effective way in improving the social perceived value but with decreasing marginal effect．However，as far as risk management is concerned，different users represent distinct preferences．In general，users with lower reference points can benefit from it while users with higher reference points will lose their potential gain．