Research On The Key Technology Of Hospital Medical Quality Comprehensive Analysis System

Hospital medical quality is related to the treatment and rehabilitation of patients and hospital's image in the social public.Medical quality evaluation is the key part of hospital medical quality management.Medical quality index is the main basis of hospital medical quality evaluation.At present there are two main kinds of methods to evaluate medical quality: Using statistical methods to make statistical analysis for one type of medical quality indexes;Medical quality multi-dimensional comprehensive evaluation method,such as the comprehensive index method,rank and analogy method,etc.Statistical methods and comprehensive evaluation method both can't make forecast analysis for medical quality index.Visible,the use of data mining algorithm to make forecast analysis research for medical quality index has obvious value of research and application.After studying a large number of domestic and foreign medical data mining and the medical quality evaluation,this thesis based on the hospital business data using decision tree C4.5,naive bayesian classifier and bayesian optimal classifier algorithms to process research about making forecast analysis for the medical quality index.Our main research work and contributions are as follows:(1)Make the comprehensive theory expounded and detailed analysis of the steps about the decision tree C4.5 and bayesian classification algorithms.First,expound decision tree,bayesian classification and the significance of the corresponding concepts,introduce the related calculation formula.Secondly,discuss information entropy,information gain,split information measurement and information gain rate in C4.5 algorithm,and maximum likelihood assumption,maximum posteriori assumption in bayesian classification algorithm,introduce the difference between naive bayesian classifier and bayesian optimal classifier.Finally,give the algorithms' steps and part of the implementation pseudo code of decision tree,naive bayesian classifier and bayesian optimal classifier,and analyze their respective characteristics.(2)Put forward and realize the prediction analysis method on the medical quality index by constructing the C4.5 decision tree mining model.Firstly,give a pretreatment on the hospital's inpatient medical homepage records,and divide the data into three datasets: training dataset,validating dataset,testing dataset.Secondly,construct C4.5 decision trees by training dataset of two medical quality indexes: average hospitalization days and diagnostic accordance rate.In addition,clip the decision tree by error rate reduce pruning method based on testing dataset,and make the classification prediction test both on decision tree model before and after clip by using testing dataset.The result shows that the decision tree model's prediction accuracy towards on average hospitalization days and diagnostic accordance rate before clip are 81.95%,82.92%,and the rates after the clip are 83.79%,85.03%.(3)Put forward the prediction analysis method on the medical quality index by constructing the naive bayesian classifier and bayesian optimal classifier mining model.Firstly,construct naive bayesian classifier and bayesian optimal classifier based on average hospitalization days and diagnostic accordance rate by using training dataset.Then do the prediction analysis testing on the classifiers by using testing dataset.The result shows that the naive bayesian classifier and bayesian optimal classifier's prediction accuracy towards on average hospitalization days are 80.63% and 84.82%,and the accuracy towards on diagnostic accordance rate are 82.01% and 86.15%.(4)Design and implement the data warehouse of medical quality and the synthesis analysis system of hospital's medical quality.Firstly,have comprehensive elaborates of data warehouse from concept,characteristic and architecture.Secondly,design and implement the data warehouse form five aspects: the way of data extraction,theme,dimension,granularity and hierarchical division,fact table and dimension table.Lastly,design and implement the synthesis analysis system of hospital medical quality,and integrate the data mining model.The thesis put forward and realize the methods of prediction analysis for medical quality index by using the decision tree C4.5,naive bayesian classifier and bayesian optimal classifier algorithms successively.Accomplish the performance comparison and analysis of the above algorithms through the experiment,The results shows that the three kinds of algorithm all can achieve higher prediction accuracy,and the bayesian optimal classifier is the highest,the naive bayesian classifier is the lowest.The above methods all have high scientific research and application value,can be used widely in the field of hospital medical quality related research.