<p>The science and technology innovation board (STIB) is a critical initiative to support the high-quality development of technology innovation enterprises. Accurate valuation of STIB-listed enterprises is essential for optimizing resource allocation and enhancing capital market efficiency. However, existing valuation methods face significant challenges, including the presence of nonlinear data and low accuracy when assessing these enterprises. Given the capability of machine learning methods to address such issues, this study proposes a hybrid approach that combines Spearman correlation analysis and XGBoost feature selection to identify key indicators. The selected features are subsequently input into a GA-BP neural network for training and simulation. The empirical results, based on a dataset of 1558 observations, demonstrate that the XGBoost-GA-BP neural network model achieves a coefficient of determination (R²) exceeding 97% and maintains a mean absolute percentage error (MAPE) below 10%. These findings indicate that the proposed model can effectively assess the valuation of science and technology innovation enterprises with high accuracy, underscoring its robust practical applicability and reliability. This study not only advances methodologies for enterprise valuation but also provides actionable insights for stakeholders to optimize resource allocation and enhance enterprise value.</p>