Mobile telecoms operators possess an enormous quantity of data, which could be used to reduce the cost of installing new infrastructure, to provide a better QoS or to plan their infrastructure. Thus, they are concerned to model, understand and predict SMS and calls activity levels in their infrastructures. Besides, SMS and call activities analysis can open new business opportunities for geomarketing as well as trade area analysis. In the present effort, we detected activity zones with a difference of only 0.5 km from the reference activity areas extracted from Geo-tweets. We also used Markov chains to represent and predict SMS and call activity levels, achieving a prediction success rate between 80% and 90%.