Researchers at the UPM have designed a system to help users of social networks on mobile devices to filter and select the most interesting information to them at all time. 


The use of social networks using mobile devices is a usual practice in the last years, therefore it requires of tools that help user to filter the huge amount of information generated by these networks.


From this idea, researchers of the Facultad de Informática at the Universidad Politécncia de Madrid have designed SOMAR, a system of recommended activities that analyze information and, by using data mining techniques, seek and reveal the relevant information to the user.

To make recommendations, SOMAR (SOcial Mobile Activity Recommender) can access to the information from three sources: (1) mobile data in terms of call history or contacts, (2) sensor data from the location and (3) data from the most used social network, Facebook. Taking into consideration privacy issues, the only information that will be used it will be the data approved by the user, and this will be the basis of the recommendation tool.


SOMAR recommends you personalized activities for every user based on the interactions with other users in the network. To do this, a social graph is calculated based on the social relationships and the common interests between friends by using clustering techniques. Figure 1 shows a SOMAR architecture and different compound modules.


Apart from the modules involved in the creation and updating of the social graph, the other modules are used to provide an appropriate recommendation by an established procedure. Thereby, while a module integrates the data from three main sources (sensors, telephone and social platform) and prepares them for further analysis, another module has as a main objective to detect activities by using a process of data mining.

Figure 1: Andrea Zanda 2011. SOMAR componets.

Although this work1 is focused on the user activities, the underlying method allows us to use a SOMAR as a basis to recommend any other type of information on a social network.


SOMAR was used experimentally with data from Facebook and the results obtained show the possibility to make personalized recommendations in real time and having into account the current context.


1 Zanda, A; Eibe, S; Menasalvas, E. SOMAR: A SOcial Mobile Activity Recommender. EXPERT SYSTEMS WITH APPLICATIONS 39 (9): 8423-8429. Jul 2012