Analyzing Customer Behavior Model Graph (CBMG) using Markov Chains

Performance is one of the main challenges in designing an e-commerce or e-business application. This article proposes a new method for estimating one important parameter of system workload, the average visit length. In order to characterize system workload, common techniques, like Customer behavior model graphs (CBMG) or Markov chains can be applied. This paper introduces a new method for transforming CBMG graphs to stable Markov chains. It is proven by measurement results that the average visit length converge to the stationary distributions of the Markov chain representations. A new method for calculating the average recurrent time as the average session length is also presented.