Electricity plays a vital role in our daily routine activities. Particularly, educational institution power requirements may vary every day and the infrastructure of the building determines the amount of power required for a day. Hence advance prediction of power for every building in a university is very important for effective management of electricity during power crisis. In this paper, prediction of power consumption of each building in a university is carried out based on the infrastructure of a building. Feature engineering is carried out using ClassifierAttributeEval. This helps in reducing the size of the dataset as well as the computation time of prediction models. To forecast the energy production and usage in advance, Support Vector Machine (SVM), Neural Network (NN), Random Forest (RF) and Stochastic Gradient Descent (SGD) models are applied. From the experimental results and 10-fold cross validation sampling type , it is proved that SGD has better prediction accuracy when compared to SVM, NN and RF with Mean Square Error (MSE) of 1.102 , Root Mean Square Error (RMSE) of 1.050, Mean Absolute Error (MAE) of 0.613 and Coefficient of determination (R2) of 1.0000.
Kalamegam et al. (Mon,) studied this question.