Cyclic pressure pulsing using CO(2) and N(2) is an effective improved oil recovery method in naturally fractured reservoirs. Determining the optimum design parameters for the process is an arduous task due to the computational cost of simulating a large number of injection schemes. In this paper, we present neural-network based proxy models that mimic a reservoir simulation model and provide estimated quantities of critical performance indicators. The proxy models are trained with a set of representative design scenarios. These design scenarios are run in a compositional, dual-porosity reservoir model and corresponding performance indicators are collected. Cyclic pressure pulsing process is modeled using two huff 'n' puff design schemes with variable and constant cyclic injection volumes. The reservoir model is constructed based on reservoir characteristics of the Big Andy Field in Kentucky which is a depleted, naturally fractured reservoir with stripper-well production. Predictive capability and accuracy of developed proxy models are checked by comparing simulation outputs with proxy outputs. It is observed that neural-network based proxy models are able to accurately predict the performance indicators including the peak rate, time to reach the peak rate, cycle flow rates, incremental oil production, and gas-oil ratio. The proposed methodology is practical and computationally efficient in structuring more effective decisions towards the optimum design of the process. (C) 2010 Elsevier B.V. All rights reserved.