A data stream is essentially a virtually unbounded sequence of data items arriving at a rapid rate. Mining frequent patterns from the stream of data is a dif-ficult task. This paper use non-homogeneous Poisson pro-cess to find the frequency of an item set from transactional data streams. We develop models by using mean value function from Goel–Okumoto model. The concept of split Poisson process and Bayesian model are used for devel-oping a model for the prediction of number of arrivals of an item with in a particular time period.