The state-of-the-art energy management for the heating, ventilation and air conditioning (HVAC) system uses a static clothing model that calculates the occupant's clothing insulation as a fixed value based on outdoor air temperature measured at a particular time of the day. However, the static clothing model can hardly capture the occupant's intra-day clothing behaviors, leading to inaccurate thermal comfort assessment and unrealistic HVAC energy management. This paper proposes a novel HVAC energy management scheme to optimally schedule the thermostat setpoints of HVAC and to provide recommendations on occupants' optimal hourly clothing decisions through a predicted mean vote model, while considering uncertainties in the outside temperature. The proposed HVAC energy management scheme is solved by applying an approximate dynamic programming approach. Further, a model predictive control framework with a long short-term memory based forecaster is developed for more realistic simulations. We study the HVAC schedules in a residential home with summer and winter time of use electricity tariffs for both male and female occupants. Compared with non-optimized cases, proof-of-concept simulation results demonstrate that the proposed scheme can achieve a 53.8% and a 29.8% cost saving in a summer-male scenario and a winter-female scenario, respectively.