To estimate the building load Based on the number of customers and stay, a new model was built with the Adaptive Neuro-Fuzzy Inference System (ANFIS) to research its feasibility and applicability. With the average error and the coefficient of multiple determinations estimating performance of the model, ultimate results showed the ANFIS model is very valid, and the accuracy rate is 99.8%. Therefore it can realize regulating and controlling compartment loads according to the number of customers and stay for market buildings. Comparing to traditional temperature and humidity control of the entire space to adjust to the change of the load, the new method can rapidly change the local environment to meet the requirements of customer comfort, and fully realize energy conservation.