Multimodal Medical Data Intelligent Classification Method and System Implementation Based on Improved SVM and Similarity Learning Algorithm

Intelligent diagnosis and treatment play a crucial role in the era of medical big data, especially in processing multimodal electronic medical record data.This article focuses on the patient similarity classification task and proposes two effective classification methods to address challenges such as poor data integrity and diverse expressions.A patient similarity learning method based on multimodal summary extraction is proposed by utilizing the GPT model for TLDR summary extraction and combining it with an improved Support Vector Machine (SVM).A patient similarity learning method based on unstructured text approximate matching is proposed using Jieba segmentation, bag of words model, and improved Jaccard function, combined with XGBoost method.Based on the above method, a patient similarity classification system was designed and implemented, providing doctors with a tool for rapid diagnosis.These methods still have shortcomings, and future work will explore lightweight network architectures, design algorithms that consider dynamically adding patient features and time series, and integrate image classification and text classification to promote further development of intelligent diagnosis and treatment.