This study presents a machine learning-based intelligent cabin alert filtering system for vehicles aiming to address safety risks caused by excessive and redundant alarm sources. To overcome limitations in current systems, such as alarm redundancy and inaccurate classifications, a hybrid selection strategy is proposed that combines manual expert filtering with a convolutional neural network (CNN) model. The system integrates operational data from various devices, applying manual heuristics to eliminate likely false signals and employing the CNN model for robust feature extraction and precise classification. Experimental results show that the CNN model achieves a classification accuracy of 89.07% on the test dataset. When combined with manual filtering, the overall selection accuracy of alarm signals reaches 99.998%, significantly surpassing the conventional VAS system (90%). These results validate the proposed method’s effectiveness in filtering alarm information. Future research will focus on expanding training datasets, optimizing model parameters, and improving text pre-processing techniques to further enhance the overall system performance.