A High-Frequency Data-Driven Machine Learning Approach for Demand Forecasting in Smart Cities

Figure 8

Forecasting at the industrial area on Wednesday, 11 March 2015. The effect of missing data in both the prediction and query days is visible in this figure. When there are missing data in the query day, the neighbors are improperly obtained since only the remaining data are considered, and thus, one might obtain a misleading neighbor which would yield erroneous forecasting results. When the data are missing in the forecasting day, there cannot accurate predictions.