Sequence Mining TV Viewing Data Using Embedded Markov Modelling

The advancement of Internet Protocol television (IPTV) has brought significant advantages to TV users in terms of the TV content viewing and control over the Electronic Program Guide (EPG) system, which has provided richer TV viewing behaviour data to gain insights from. The Set-Top-Box (STB) in an IPTV system typically collects user behaviour data as well as monitors system performance in the form of event logs. Sequence mining technologies have been widely used in the study of event logs, such as mining the most frequent items. This paper presents the analysis of event logs on an IPTV box (BT YouView Box), focusing on customers’ navigation paths on the EPG system. We propose a new methodology by utilising embedded discrete-time Markov Chains (DTMC) and dynamic programming, to generate maximum likelihood trajectories for customer’s navigation sequences. This paper has demonstrated that the DTMC fits well into event logs and has revealed insights into customers’ navigation paths. The analysis has also identified several potential areas for further work.