A Theoretic Framework Integrating Text Mining and Energy Demand Forecasting

News articles are an important source providing information about society. The analysis of news articles helps to measure the social importance of many events and to give an understanding about current interests. In this research, a theoretical framework of text mining enhanced approach is proposed to accommodate short-term variations caused by special events, such as severe weather conditions. A sentiment analysis approach for extracting sentiments associated with positive or negative polarities from a series of news reports is utilized to illustrate impact on energy demand from a special event. The magnitudes of the sentiments from the series of news articles are used to compose a timeseries pattern to represent the event that translated into the causes of short-term demand or price variation. The proposed approach for sentiment analysis is demonstrated with experimental results. In particular, the development of an event pattern using text mining is illustrated. The implications of the proposed framework and the future research direction are discussed.