Suppose we have N sources each leading to M processes. Each process is quantified by a probability of occurence. The probabilities of occurence for M processes corresponding to N sources are available for 5 weeks. What is the best possible approach to detect anomaly in these processes using the probabilities?