It is to be noted that the division of data for training and testing purpose is of the ratio of 3:2. Such high prediction rate is attributed to the improved clarity of the data set that is fed to the neural network for training. Rather than providing crisp energy consumption values, the class numbers represent a better way of classifying energy data and make it easier to comprehend the existing pattern of energy use from one day to another. Fig. 6 shows the plot with
measured and predicted values for building A using ANN. It is observed that the slope of the line is also close to one.
Similar results for building B and C are presented in Table 3.