An algorithm based on Google Trends' data for future prediction. Case study: German elections

The analysis of the high volume of statistics generated by web search engines worldwide on a daily basis, allow researchers to examine the relation between the user's search preferences and future facts. This analysis can be applied to various areas of society such as sales, epidemics, unemployment and elections. The paper investigates whether prediction of election results is possible by analyzing the behavior of potential voters before the date of the elections. In particular, the proposed algorithm is applied on the three more recent German elections. The results of this analysis show that a strong correlation exists between the search preferences of potential voters before the date of the election race and the actual elections results. It also demonstrates the fact that search preferences are influenced by various social events that may take place concurrently to the election race. The effect of such events has to be filtered out as noise in order to arrive at a successful estimation of the final results.