A method of model reduction using the universal maximum entropy algorithm in conjunction with Norton's theorem for queueing networks is proposed for the analysis of large G-type distributed closed queueing networks. This method is called Norton-Maximum Entropy (N-ME) and has an advantage over the direct application of universal maximum entropy whereby the parametric study of a subset of queueing centres of interest can be done repeatedly without solving the entire network. The complexity of the original system is reduced into a flow equivalent two-stage load dependent queue. The marginal probability density function and various performance parameters of the two-stage load dependent queue can be obtained by using the maximum entropy analysis of the GE(n)/GE(n)/I/Nqueue.