To address the challenges associated with effectively implementing dust reduction measures in the fully mechanized excavation face, this study introduces a rapid flow field prediction algorithm based on proper orthogonal decomposition (POD) and machine learning. Initially, the algorithm employs POD to perform a dimensional reduction of flow field data or dust concentration data across multiple working conditions, yielding the basic function modal and modal coefficients of the flow conditions. Utilizing machine learning techniques, the algorithm predicts the modal coefficients for diverse conditions that contribute to over 90% of the total energy, facilitating the estimation of modal coefficients for unknown scenarios. Reconstructing using these predicted modal coefficients alongside basic function modals enables the derivation of flow field data or dust concentration data for unknown conditions. Comparative analysis reveals that the Support Vector Machine (SVM) model exhibits superior capability in predicting modal coefficients compared to other models. Using numerical simulation results from 300 conditions in the comprehensive excavation as data samples, the SVM model can predict the flow field and dust concentration field in just 1/816th of the time required by traditional numerical simulations. Additionally, a comparison between the predicted results and numerical simulations across 60 conditions indicates a relative error of 0.36m/s for flow velocity and 86.24mg/m3 for dust concentration across all grids.