An Intelligent Forecasting Model for Building Energy Consumption Using K-shape Clustering and Random Forest

Intelligent building energy consumption is a typical unbalanced dataset. Workdays take up high proportion so the features of workdays are more likely to be taken into consideration while the features of the rest days are not learned fully even filtered as useless information. Therefore, RF model is not able to achieve ideal performance in intelligent building energy consumption forecast. It need some else technology help to deal with unbalanced datasets. Time series clustering can make up the disadvantage and k-shape is an accurate clustering algorithm with low calculation cost. k-shape excavates data features so that several basic energy consumption patterns are identified and noisy data is labeled as well. This extra information help determine the possible pattern which serve as reference, reduce the negative effect of noise in the meanwhile. The accuracy of forecast model is improved in this way.