The economic operation indices of industrial boilers are influenced by multiple factors. Multi-objective optimization is employed commonly to identify optimal operating conditions under multiple objectives and constraints. However, the obtained Pareto Front typically consists of a large number, offering diverse trade-offs and requiring a proper data selection for application. To address this challenge, the operating dataset was collected from a fire-tube steam boiler using natural gas as a fuel with a capacity of 22 ton-steam/h. A principal component analysis (PCA) was applied to reduce the dimensionality of both the operating dataset and the Pareto Front dataset into a 2D principal component space. Each operating point was then paired with the most similar Pareto-optimal point using the shortest Euclidean distance (i.e., 2D-Nearest), subjected to a specified deviation in operating load level. The pairing results were further evaluated against two single-criterion methods the minimum gas flow and the maximum combustion efficiency. The comparison indicated that the optimum dataset constructed using the 2D-Nearest method was the most outperformed approach by achieving the smallest specific fuel consumption (SFC) of 47.8 kg-gas/ton-steam while increased in combustion efficiency of 86.47%. In addition, an analysis of distance distribution showed that the proposed method provided the smallest mean distance, indicating a closer resemblance to the boiler actual operation compared to the rest pairing methods.