Medical registration is the most basic unit of the medical profession. Generally patients don't understand the condition of their illness. So choosing the wrong department is completely common. Intelligent medical system solves this problem very well. Intelligent medical system makes use of the massive medical record texts which the medical department accumulates to train and carry on machine learning, and to analyze the characteristics of the registration patient's medical record and to classify to the right disease. The patient is recommended to the appropriate department according to getting the department to be registered. The influence on the efficiency and accuracy of the traditional support vector machine (SVM) text classification of the intelligent medical system is the feature extraction and kernel function parameters optimization. Therefore, the method of the Genetic Algorithm (GA) combining with SVM is proposed. The text feature values and kernel function parameters together is viewed as a chromosome of the genetic algorithm that is an individual to carry on the binary encoding. The optimal individual is obtained by the genetic manipulation of the selection, crossover and mutation. Finally, text classification is operated by support vector machine using the optimal features and optimal parameters. The test result shows that this model improves the accuracy of intelligent diagnosis and is a good intelligent diagnosis registration algorithm.