Paper Title
An Analysis for Feature Selection on Hourly Electricity Load Forecasting

Abstract
The increasing energy needs get more importance to different energy sources. The increase in the diversity of energy resources has mandated the proper management. Prediction of the load demand is an important part of the energy management issue. The accuracy of a prediction model is directly related to features used in the model as inputs. Therefore, feature selection is a way to increase prediction performance. In this study, hourly load demand of main campus area of Afyon Kocatepe University is used. The season, month, day and hour of the consumption are considered as inputs besides the past consumption value. The impact of the features on load forecasting accuracy is investigated based on artificial neural network. The analysis show that hour and day of the consumption features carry more information than season and month. Keywords - Load Forecasting, Artificial Neural Network, Feature Selection