In view of the low efficiency of manual analysis when dealing with massive medical records in clinical practice, this study constructs an AI-powered intelligent analysis and decision support system for TCM acupuncture based on multimodal data mining. The system integrates multiple functional modules to enhance diagnostic accuracy and treatment efficiency. The multimodal data fusion module leverages an autoencoder model to integrate tongue images, facial images, and pulse information into a comprehensive feature vector, combining key textual descriptions to generate a holistic patient health profile. The feature extraction and selection module applies decision trees (DT) to identify the most critical features for acupuncture treatment decision-making. The intelligent analysis and decision support module employs a support vector machine (SVM)-based prediction model to process the extracted features and generate personalized acupuncture treatment plans. The user interface and interactive module enable seamless data entry, treatment recommendation visualization, and historical query access, ensuring ease of use for physicians. Experimental results demonstrate a significant reduction in diagnosis and treatment time compared to traditional methods, while achieving a treatment accuracy exceeding 98%. The system also outperforms conventional approaches in symptom improvement rates and quality of life scores, with symptom improvement exceeding 98% in numerous cases and the highest recorded quality of life score reaching 10. By integrating advanced data mining techniques and intelligent decision support, this system not only enhances the efficiency and accuracy of TCM acupuncture therapy but also provides a scalable framework for future AI-driven clinical applications.