Optimizing energy efficiency, boosting occupant comfort, and strengthening overall building management have all made the use of smart building technologies more and more crucial. Smart buildings monitor and manage several systems, including lighting, heating and cooling, security, and energy consumption, in real-time, using sophisticated sensors, Internet of Things devices, and data analytics. Significant financial savings, less of an adverse influence on the environment, and more efficient use of resources result from this. To control the temperature of a two-zone smart building using machine learning model predictive control (MPC), this study proposes a new model predictive control approach. In this research, keeping residents comfortable and minimizing energy savings are the main objectives. In addition, model-based control is incorporated into the proposed control scheme to achieve the desired goals. To predict indoor temperature, the artificial neural network (ANN) utilized to estimates occupancy profiles in building zones over the long term, and this data is fed into the model-based predictor. As the actual data set provider, Python and MATLAB softwares are used. A set-point temperature is generated for the heater/cooler by solving the optimization problem, using both actual and predicted data. Compared with conventional MPC results, the proposed approach significantly saved energy compared to the conventional method. In the end, a practical approach example has been demonstrated to presents the effectiveness of our proposed approach.