Data Cleaning for Prediction and its Evaluation of Building Energy Consumption

Building consume a large amount of energy and a plenty of methods to mine into energy consumption data to aid intelligent management are proposed. However, the data quality issues are inevitable and the influence is lack of discussion. This paper proposed a data cleaning method combing threshold and cluster method. This paper also proposed an index to evaluate the accuracy improvement on big data prediction. A case study is conducted and it is found that the accuracy of data filling is not sure to agree with the improvement of prediction after filling.