<p>In this paper we propose the affinity model for classifying social groups based on mobile phone call detail records. We use affinity to measure the similarity between probability distributions. Since phone calls are stochastic process, it makes more sense to use probability affinity to classify the social groups. This work is useful for enhancing homeland security, detecting unwanted calls (e.g., spam) and product marketing. For validation of our results, we used actual call logs of 100 users collected at MIT by the Reality Mining Project group for a period of 8 months. The experimental results show that our model achieves good performance with high accuracy.</p>