| With gradual development of information science and technology,the digitalization level of traditional industry has been continuously growing.Researchers in healthcare filed have made great progress in recent years to break through the barriers between healthcare and information technology.Business intelligence technology,as an important tool of industrial digitalization,can help healthcare practitioners to integrate existing large scale of data,reorganize business logic,and implement deep data mining.On one hand,in current Internet environment,healthcare big data and data mining can be used to discover the trend of disease outbreak across the world,help establish personal health records and optimize individual daily health management;on the other hand,business intelligence can be used in hospital operation and service,with great help to solve resource optimization problems derived from doctor shortage and large amount of population.With such two directions,4 research points are proposed as following.(1)For healthcare data on Internet,in the background of online medical Q&A community,to solve the problem that users cannot get timely and effective answer,users’ interest of filed(IoF)model is proposed to represent users interest to different fields based on query likelihood language model,with utilizing user historical activity data;at the same time,this paper propose users’ personal expertise model to represent individual expertise in response to a professional medical problem.With above two models,this paper presents a weighted user model that can be used to predict if a user can answer questions timely and effectively.(2)For offline data in healthcare field,especially patient information data included in electronic medical records,due to the unstructured characteristic,it is difficult for readers to extract key information in a short time.A key information extraction algorithm for electronic medical records is proposed.With clear definition of key information in electronic medical records,medical natural language processing and neural networks are utilized to build key information extraction algorithm based on supervised method.After that,word vector characteristics model is established based on deep neural network model,which is integrated to final key information extraction algorithm.(3)To solve the problem that emergency department is visited too frequently with unnecessary encounters,and to optimize emergency service for hospitals,in this paper patient historical visit activity data is utilized to build an emergency patient admission prediction algorithm based on random forest model and ensemble modeling.Through statistical learning,key risk characteristics to determine future emergency visit are discovered.After complete risk assessment,a high risk patient clustering analysis algorithm is proposed to identify potential medical patterns,for the purpose of helping providers perform early medical intervention for high risk patients.(4)At the end of this paper,based on previous research work,a complete set of business intelligence solution is proposed to implement deep data mining for electronic medical record data.An analysis platform for whole workflow process is constructed,with a universal healthcare data warehouse as foundation,an integration process for cross-platform data as workflow,and enterprise dashboards based on key performance indicators as data visualization interface. |
