Article Preview
Top1. Introduction
Natural gas is one of the major energy resources in the world. It has been widely utilized in a variety of aspects, such as residential, commercial, industrial, and power generation. Natural gas is a commonly used fuel for residential cooking and heating, and an essential energy material for commercial electricity generation (U.S. Energy Information Administration, 2013). Therefore, the timely and accurate prediction of natural gas prices is of great importance. The natural gas price is determined by many complicated factors reflecting its demand and supply. Existing studies on natural gas price forecast consider traditional statistical data source like production, storage, import and export of this resource, as well as economic growth, oil price, and even weather (e.g., Buchanan et al., 2001; Mu, 2007; Brown and Yucel, 2008). However, little effort has been devoted to applying massive user data from Internet, which seem to be a more direct way to represent natural gas demand. To fill this gap, this paper puts forward a new perspective by incorporating user search data generated from Internet to represent public attention in order to improve natural gas price forecast accuracy.
In the Big Data Era, public attentioninformation established from Internet-based knowledge has become a new influencing factor on price for the following two reasons (Barber & Odean, 2008; Da et al., 2011). Firstly, the price variation of natural gas is closely related to the public’s daily life. Both residents who need gas to cook and keep warm and the business managers who need it to keep operation of the company should pay attention to natural gas price. Secondly, the public attention represents their demand dynamics, which determines the natural gas prices in turn (Krichene, 2002; Huntington, 2007). Accordingly, to investigate whether public attention could improve the forecast accuracy of natural gas prices is an innovative and meaningful research topic.Due to the development of Internet technologies and massive data processing methods, it is possible to timely collect the public behaviour data. Thisused to be difficult because seldom data sets provided timely public behaviour information in the past. Nowadays, people carry out many online activities, such as searching information, expressing opinions, and communicating with others through social media platforms. Besidesbehaviour data like search keywords, page views and the re-tweets,some personal information that is not conflict with privacy policy are recorded by software. These online data contain abundant knowledge and couldpossibly reflect the public demand.