An evaluation model of core enterprise performance was proposed based on rough sets and neural network from knowledge discovery and data mining perspective at first. Then, the performance decision-making table and discernable matrix were designed and the neural network and back propagation algorithm (BP neural network) were put forward. Finally, the model was applied into performance evaluation study of a manufacturing core enterprise. After the index system of the core enterprise based on the balanced scorecard method was reduced and the reduction index was input to the neural network for intelligent training, the evaluated sample of the company was input to the trained network, the evaluation value of the manufacturing core enterprise performance was gained. The result indicates that the evaluation result is consistent with the actual result.