A General Purpose Model for Future Prediction Based on Web Search Data: Predicting Greek and Spanish Election

Abstract

Although real time data processing consumes a significant portion of computing resources worldwide, we are moving rapidly from the age of "real time" towards the era of "next time". The term "next time" characterizes the combination of real time data flows from collective sources with massive computing power with the aim of predicting the future. In other words, if you can compute "fast enough" using real time data sets then you can accurately predict what happens next. Central to this process is the Google Trends service that provides generalized statistics in regard to the popularity of web search terms submitted to Google. The paper combines the conclusions derived from other approaches to the prediction problem with Google Trends data in order to predict the outcome of six national elections races in two countries, Greece and Spain. The results of the proposed model reaffirm our hypothesis that web search terms popularity is directly related to voters decisions in both countries and thus can be used to predict the final outcome with great accuracy.