In the recent cyberspace environment, the intensive and extensive use of search engines is commonplace. Tracking usage patterns of users and then aligning with the implicit user requirements is an essential competition factor among search engine manufacturers. There are several approaches for positive feedback to monitor the impact of freshly introduced changes on users. This paper proposes a mathematically grounded method for analysis of user behavior. The research combines the discrete Fourier transformation that is used for the analysis of frequencies of user activities and traditional statistical methods, for example, cluster analysis.
The method is applied to investigate the effects of modifications in a specific search engine. The researchers use the log file of a search engine and therefore have a statistically significant set of data.
Frequency or periodicity can be examined with the help of the discrete Fourier transformation method, which provides the transformation between the frequency and spectrum domain. The transformation allows for the application of cluster analysis whereby the typical patterns of user behavior can be explored.
The conclusion of the research is that the introduced periodicity metrics along with the proposed analysis method are much better than previously used basic/naive metrics. The paper is interesting for researchers and professionals interested in social networks, search engines, and text and data mining methods.