The technology features demonstration of the next generation fighter is a significant issue for increasing military equipment performance, leading technology developments and supporting specific single weapon demonstration. This paper introduces data science to compare traditional backward induction methods with machine learning, and points out the weakness of traditional methods during the technical features demonstration of the next generation fighter, such as the sample dependent, delay, unexplainable and overfitting. Then a mathematical programming based demonstration method is proposed. This method obtains technology features of the next generation weapon with optimal overall cost, which is subject to capability requirements such as mission and environment requirements. Finally, the proposed method on US Navy and Russia Air Force sixthgeneration fighters is evaluated to prove its feasibility and superiority.