This paper adopts data processing methods and data mining technology to develop a building energy consumption model, based on an energy consumption database of commercial buildings that includes 95 commercial buildings in Shanghai. Data transformation and data reduction are conducted to clear up data relations in the database. Three methods for missing data handling as well as outlier inspection are used for data processing. The software SAS is used as the tool for data processing and data mining. An optimum regression model of building energy consumption is made for each missing data element. Through comparing the three optimum regression models and their prediction results of building energy consumption, it is found that the Regression Imputation Method was the best method to handle missing data, and a regression model with operation time of HVAC system, cooling capacity, ratio of office area to total gross area, and hotel area to total gross area was the most reasonable prediction model of the energy consumption of commercial buildings in Shanghai.

Units: SI