I don't know how many of these factors are used, but I'd do a statistical analysis of many of these factors, in addition to what slawecki listed:
Historical frequency FOR THIS FLIGHT of last minute (last few days?) no-shows/cancels/transfers to other flights.
Other route alternatives(IE: last flight of day for this route may have lower no-show rate as no other option that day, while routes with many flights/day have higher no-show/change rate earlier in day. Unless, of course, stats shows that many pax with last-flight-of-day reservation go standby on earlier flights, then can overbook last flight at higher frequency)
Percent of flyers with refundable fares (higher last-minute cancel rate?)
Percent of direct vs. transfer pax. (If pax is ending travel after this flight, lower % of no-show or change to other route on day of travel or soon before. If only a flight to a hub where most pax transfer to another flight, higher % of last-minute route or time change, etc.)
Percent of flights that are cancelled for this route. (If one of 10 first-flight-of-day-for-this-route is cancelled, need to leave more seats open on next flight of day.)
etc...