The Research Of Intelligent Assistant Diagnosis Model For Short Stature

Short stature,a common disease in endocrinology of Chinese children,has hurt 8 million children in our country,which not only causes serious psychological illness for the patients,but also affects their work and life.The increase of morbidity will aggravate the society economic burden that root in the help for dwarfism patients,therefore,it is particularly important to timely diagnose and treat the patients.However,due to the complexity of aetiological agent,plenty of doctors cannot accurately diagnose the definite causes.The model establishment of short intelligent assistant diagnosis shows excellent value,it will alleviate the burden on the expert doctor via predicting the dwarfism etiology and further provide the early diagnosis and treatment to remote areas.This paper provides decision support for the diagnosis of dwarf causes.Intelligent assistant diagnosis model for short stature is established through machine learning.The main research contents of this paper are as follows:(1)To learn the theoretical knowledge about subject in the context of Short stature,including the definition and pathogenesis of short stature,and the diagnostic methods and procedures of doctor.At the same time,the theory knowledge in the experimental process is studied,including Chinese segmentation,classification algorithm of machine study and model evaluation indicators.(2)Data preprocessing.This step is to transform electronic medical records from semi-structured text into structured data.First,the electronic medical record is parsed,the text information needed is extracted.then,texts are segemented,and dictionaries and stop words are added.Next select candidate features by calculating tf-idf.Finally combine the expert knowledge to get the final features.(3)Model building.The C4.5 algorithm was used in the first experiment.In order to improve the generalization ability of the model,the second experiment uses the Adaboost algorithm.For the problem of unbalanced positive and negative samples in the experiment,the third experiment uses the Adaboost algorithm based on the improved sampling method.(4)Model evaluation.Compare the three models by true positive rate,true negative rate,and accuracy.The experimental results show that the performance of the Adaboost algorithm based on improved sampling method is greatly improved,which classification accuracy is 85%.Therefore,the optimized model is more helpful in providing decision support for short diagnosis.