Study On The Risk Prediction Model Of Elderly Patients Before Operation Based On Cluster Learning

With the acceleration of the aging process of China’s population,the demand of elderly patients for medical treatment is increasing,and the corresponding number of elderly surgery cases is also increasing year by year.Although the current medical level and technology are developing rapidly,the risk of surgery still exists.Especially for the elderly patients,because of their physical quality accompanied by a large number of chronic diseases,the risk of surgery is greatly increased,and the surgeons for the elderly patients are often very cautious.However,the elderly patients have higher and higher expectations of longevity,and they are more and more willing to improve their quality of life through surgery,so it is particularly important to do a good risk assessment before surgery.Most of the traditional surgical risk assessment systems only provide a simple probability prediction,and the evaluation method is easily affected by subjective judgment,and the prediction information provided is often very limited.However,most of the surgical risk assessment systems of many individual studies are aimed at single disease or specific population.The existing systems often fail to provide sufficient decision support ability for clinicians during the preoperative preparation.At the same time,there is a gap in the generalpurpose surgical risk prediction system.If the surgical risk prediction system can provide the medical records of similar patients,so that doctors can understand the relevant surgical risks as early as possible,and be prepared,it can greatly reduce the probability of risk.In order to solve the shortcomings of the traditional risk prediction system,this paper proposes a clustering based surgical risk prediction model for elderly patients.Based on the study of the elderly surgical patients in a large-scale second-class hospital in Shanghai,the relevant data of patients over 60 years old in the past two years were selected.According to the mode of clinical preoperative consultation,the sample features to be studied are divided into surgical features,test features and preoperative summary.The operation features and test features are clustered in blocks,the optimal model is selected and the learning samples are labeled.When forecasting the samples,we use Chinese text clustering to calculate the similarity,combined with the label of decomposition features,and get the results of risk prediction by rules.Compared with the existing models,this paper provides a new research idea for the research of general preoperative risk prediction model,which can provide risk prediction and recommend similar risk medical records at the same time,greatly improving the availability of surgical risk prediction model.An improved k-means algorithm is proposed for clustering analysis of features.Through the combination of SOM algorithm,the final weight of its competition layer nodes is taken as the clustering center of K-means algorithm.Combined with the information gain algorithm and the improved relief algorithm,each feature dimension is given a weight,and the influence of different feature dimensions on the clustering results is considered.At the same time,combining the contour coefficient and elbow method to determine the range of initial clustering number,the traditional K-means algorithm is improved.The algorithm can be more in line with the research of clinical characteristics.Finally,the experiment also proves that the improved kmeans algorithm has higher accuracy and better effect than the traditional K-means algorithm.