Using Markov Models to Define Proactive Action Plans for Users at Multi-viewpoint Websites

Deciding about the best action plan to be tailored to and carried out for each user is the key for personalization. This is a challenging task as the maximum number of elements of an environment have to be taken into account when making decisions such as type of user, actual behaviour or goals to fulfil. The difficulty is even greater when dealing with web users and when decisions have to be taken on-line and salesmen are not involved. In this paper, we propose an approach that integrates user typologies and behaviour patterns in a multidepartamental organization to decide the best action plan to be carried out at each particular moment. The key idea to do this is based on detecting users behaviour changes by means of Behaviour Evolution Models (a combination of Discrete Markov Models). Besides, an agent based architecture has been proposed for the implementation of the whole method.