Purpose: In order to classify different types of health data collected in clinical practice of hernia surgery more effectively and improve the classification performance of support vector machine (SVM). Methods: A prospective randomized study was conducted. Sixty patients undergoing hernia repair under general anesthesia were randomly divided into two groups, PLMA group (n = 30) and ETT group (n = 30), for airway management. Heart rate, systolic blood pressure, diastolic blood pressure, mean arterial pressure, respiratory parameters and the incidence of complications related to ProSeal laryngeal mask airway (PLMA) and endotracheal tube (ETT) were collected in clinical experiments in order to evaluate the operation condition. On the basis of this experiment, at first, expert credibility is introduced to process the index value; secondly, the classification weight of the index is objectively determined by the information entropy output of the index itself; finally, a comprehensive classification model of support vector machine based on key sample set is proposed and its advantages are evaluated. Result: After classifying the experimental data, we found that SVM can accurately judge the effect of surgery by data. In this experiment, PLMA method is better than ETT method in xenon repair operation. Discussion: SVM has great accuracy and practicability in judging the outcome of xenon repair operation. Conclusion: The proposed index classification weight model can deal with the uncertainties caused by uncertain information and give the confidence of the uncertain information. Compared with the traditional SVM method, the proposed method based on SVM and key sample set greatly reduces the number of samples that misjudge the effect of samples, and improves the practicability of SVM method. It is concluded that PLMA is superior to the ETT technique to hernia surgical. The idea of constructing classification model based on key sample set proposed in this paper can also be used for reference in other data mining methods.
Keywords: comprehensive classification model; data mining; index classification weight; information entropy; support vector machine.