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

Figure 10

Forecasting at the high-density population area on Wednesday, 18 November 2015. The water flow measurements show a very large random component that increases the variance by a great amount. However, the general pattern of the series was well predicted by the forecast. Therefore, the RMSE remained low (RMSE = 2.8 l/min), even when the FOB was high (FOB = 0.51).