DRGs Grouping System Of Tumor Big Data Based On Rule Mining

The problem of imbalance between medical insurance accounts has been caused,owing to the aging of population and high incidence rate of cancer.In this paper,we propose a solution of DRGs grouping system based on rule mining.The system is based on spark on yarn distributed platform and spring boot framework,integrating machine learning,big data analysis technology,drools rule engine,natural language processing,regular expression and other technologies.According to the grouping specification of CHS-DRG related to national medical insurance disease diagnosis,Study the key issues and technologies involved in DRGs grouping.First of all,this paper analyzes the development status and version change of DRGs at home and abroad,and then puts forward the research objectives and contents of this paper.Secondly,study the related theory and technology.Thirdly,according to the user requirements and CHS-DRG grouping specification,analyze the requirements of the system.Fourthly,the overall design of DRGs group system,including the overall architecture,technical route,system function module,database design.Fifthly,it introduces the design and implementation of key modules of the system,including tumor data preprocessing based on spark on yarn,tumor intelligent coding method based on drools rule engine and text similarity matching algorithm,model comparison and selection based on regression analysis algorithm.Finally,the coding results and grouping results are tested and evaluated.The main work of this paper is as follows.1.Based on the characteristics of spark parallel computing,regular expression and Lagrange interpolation are used to clean the original tumor data,with the methods of one-hot encoding and normalization,after the intelligent coding of tumor data for data transformation.2.The tumor data after data cleaning is combined with the requirements of tumor intelligent coding,based on the drools rule engine,the main diagnostic selection rules are established,diagnosis name and operation name after rule processing,they are respectively in line with the national medical insurance coding standards: ICD-10 disease diagnosis code and ICD-9-CM-3 operation code,based on the TF-IDF method of gensim and jaro method of Levenshtein,the coding results with high accuracy are returned after text similarity matching,and then the coding with low accuracy is modified manually.3.The tumor data after intelligent coding,according to the CHS-DRG grouping scheme of national medical insurance disease diagnosis.Firstly,according to the main diagnosis of tumor patients,the major diagnosis category are determined.Combined with the main operation and gender to determine the adjacent diagnosis related group.Finally,considering the individual characteristics of the patients,such as complications and the way of leaving hospital,we get DRG coding group.4.First of all,we use machine learning regression analysis algorithm to establish the DRGs code and the relationship between the individual characteristics of cancer patients and the cost,and then predict the total cost of hospitalization.Secondly,observe the distribution of the predicted total cost,and calculate the CV of the same DRGs group,the final DRGs grouping results are obtained.Finally,based on the spring boot framework and Vue framework,develop the data service interface and interface of tumor coding and DRGs grouping.