This study presents a Design of Experiments approach to construct a proxy model which produces optimum well spacing. Profitability Index (PI) is the main objective function of the model which represents the feasibility of well spacing design. The model is specifically developed for a reservoir which applied CO2 injection and 5-spot injection pattern. The study focused on understanding the economic and technical aspects of CO2 flooding to generate proxy model and optimum well spacing. There are several parameters used to generate the model, which influence production performances. Literature study and comprehensive understanding of CO2 flooding is necessary to construct a good model. By utilizing CMG-CMOST, 3582 experiments were automatically generated which applying Latin hypercube sampling. To improve the model quality, quality control is applied which considering statistical knowledge and specific constraints. The proxy model developed a polynomial regression which gives a quite good value of R-Square, 0.9067. Furthermore, another method is applied to the model constructed by CMOST to increase the R-square value and to decrease the error. An AI-based statistical method is used which gives an increased R-square to 0.9303. Then, neural network method is applied with a 16-9 as its neural network architecture which gives the best R-square of 0.9515 and the mean absolute error of 0.28. By applying Newton Raphson method, optimization is conducted by applying inversion process to selected model which set PI to maximum by modifying the well spacing. To validate the result of optimization, sensitivity analysis is conducted varying horizontal permeability and reservoir thickness.