Modern devices typically generate large quantities of cheap log data. In order to extract relevant information from this historic data, however, a carfully chosen approach is required. In his presentation, the author will give an overview of his log mining activities in two different domains using machine learning techniques.
1. Mining of server log data - or clickstreams - with the aim of characterizing the navigational pattern of internet users (conducted at Fraunhofer IPSI)
2. Print log data - with the objective to predict failures of printers (conducted at Xerox Research Centre Europe)