In order to solve the problems of low efficiency and insufficient accuracy in the performance evaluation of traditional enterprise skills training, the author proposes a performance evaluation system for enterprise skills training based on big data technology. This system integrates big data collection, storage, and processing technologies, combined with BP neural network algorithms and data visualization tools, to achieve deep analysis and dynamic evaluation of employee training data. The system leverages the self-learning capability of the BP neural network and integrates nonlinear fitting abilities, enabling it to effectively process complex training datasets, reduce evaluation bias, and improve decision-making accuracy. The system can accurately identify key performances during the training process and effectively predict the potential for employee skill development. The research results indicate that the system significantly improves the scientificity and real-time performance evaluation of training, providing strong support for enterprises to optimize training resource allocation and management decisions. Meanwhile, the application of this system has accelerated the digital transformation of enterprise human resource management. In the future, to enhance the adaptability of the BP neural network, research will focus on optimization through the integration of deep learning techniques, thereby further improving the system's predictive capabilities and enhancing its scalability across different industries. With the further development of intelligent algorithms and dynamic data analysis technology, this system will play a greater role in applicability and flexibility, helping enterprises enhance their core competitiveness.