| Nosocomial infection refers to the infection acquired by patients in the process of hospital diagnosis,mainly including two types: 1)infection acquired by patients in the process of hospitalization;2)infection acquired by patients in the hospital but showed the corresponding symptoms after discharge.The detection of nosocomial infection is one of the important contents of nosocomial infection management in hospital,and it is also the difficulty and emphasis of nosocomial infection control in hospital.The most critical step in the diagnosis of nosocomial infection is to determine whether the patient is infected.Due to the type of infection is complex and the clinical manifestations of some infections are very similar,the distinction is difficult.Therefore,medical diagnostic decision support systems for nosocomial infection that can support the diagnosis of a patients’ infection become more and more important for healthcare professionals.Electronic medical record(EMR),which uses electronic devices,such as a database in a computer or a cloud space,to store all medical records in a patient’s paper medical record,and can quickly and accurately obtain patient information through a modern data access interface.With the continuous promotion of electronic medical records,a large number of electronic medical record data has been accumulated,and text classification algorithms have always been the focus of research in the field of data mining and machine learning.In this paper,the electronic medical record data of different infections are processed and analyzed.The machine learning method is used to model the relationship between the symptom composition,symptoms and infection of different infections.Through a series of model training methods,the intelligent diagnosis of infection with electronic medical records is obtained.The system obtains the infection that the patient may have by inputting the patient’s medical record information.The main work of this paper is as follows:· A data preprocessing flow that is suitable for electronic medical records is presented.Firstly,we need to filter the negative phrases in the text of medical records.Secondly,in order to improve the quality of word segmentation,we construct a professional dictionary containing common Chinese diseases and drug names.Finally,the segmentation results of the two special grammatical structures of noun +quantifier and noun + adjective are processed and recombined.· A feature selection method based on category discrimination is proposed.The method calculates the representativeness of each symptom for different infections according to the three characteristics of the composition and distribution of symptoms in the medical text,and selects the symptoms with large representativeness according to the ranking.The experimental results show that the feature selection method based on category discrimination is more effective than the traditional filter feature selection method,such as chi-square test.· An infection diagnosis model based on multi-label classification algorithm is proposed.Firstly,the correlation between different infections is determined,and then the appropriate multi-label classification algorithm is selected according to the relevance to construct the classification model.In order to verify the performance of the classification model,this paper used the doctor experience and infection knowledge of XY hospital to construct an infection diagnosis expert system,and compared the performance difference between the two systems.· In order to analyze the applicability of feature selection method and intelligent diagnosis model,this paper uses the medical records data of XY hospital and ZD hospital to conduct feature similarity and model migration ability,and verify the applicability of feature selection method and intelligent diagnosis model.At the same time,the quality differences of different infection data between the two hospitals were compared. |
