| With the continuous improvement of China’s medical system and the rise of joint hospitals and private medical institutions,the quantity and quality of the medical industry have been improved.However,under the implementation of the graded diagnosis and treatment policy,the medical system still has the situation that high-quality medical resources are too concentrated in large hospitals,which leads to the low medical energy efficiency of primary medical institutions,and makes patients care for primary medical institutions.The level of distrust and the asymmetry of doctor-patient information make patients unable to choose a doctor according to their condition.The existence of these problems makes patients choose a large hospital as their first choice.There are strong blindness and irrationality in the choice of doctor,which results in the aggravation of the burden of patients’ medical expenses.At the same time,the emerging network medical platform can not effectively provide the essence for patients’ diseases.Quasi-guided medical service is not conducive to solving the social problems of "difficult and expensive to see a doctor".Therefore,thinking about how to provide precise guidance and personalized recommendation services for patients is an urgent problem for hospitals,network medical platforms and government departments.Based on this,this paper studies data-driven precise guidance and personalized recommendation strategies from the perspective of patients.Based on a large number of literatures,domestic and foreign scholars have used data mining methods to conduct in-depth research on referral.However,in the current trend of hierarchical diagnosis and treatment system,there is no specific solution strategy for a specific disease matching recommendation mechanism.In order to reduce the blindness and cost of patients’ choice of medical treatment,this paper takes hypertensive complications as an example,takes doctor-patient matching as the main research object,takes multivariate data mining as the basic research method,and solves the two key problems proposed in this paper: personalized hospital recommendation and accurate doctor matching.Firstly,we use word frequency statistics,co-occurrence analysis and social network analysis in text mining to identify disease risk factors.Then,after data preprocessing and risk factors discretization,we use K2 algorithm to construct a Bayesian network risk prediction model based on disease,and then establish a recommendation table for hospital visits based on risk rank assessment.At the same time,we use questionnaire analysis to identify the risk factors.Patient’s preference for influencing factors of hospital selection,and then establish personalized hospital recommendation strategy based on comprehensive factors;secondly,on the basis of identifying doctor-patient matching characteristics,a one-way doctor-patient matching model based on Naive Bayesian algorithm is constructed and evaluated.The conclusion of the study can be used to recommend doctor categories for patients(specialist outpatient physician or nonspecialist outpatient physician).At the same time,the optimal data set of doctor-patient bidirectional matching is identified based on word segmentation function,and the model of doctor-patient bi-directional matching based on binary tree support vector machine is trained and constructed on the optimal data set,and the model is evaluated.The research conclusion can be based on the recommended doctor category,which can recommend the doctor who is most suitable for the patient’s condition in the doctor category.Secondly,taking a single patient as an example,this paper conducts an empirical study on the whole process of the precise guidance and personalized recommendation strategy established in this paper.Finally,according to the conclusions of the study,this paper puts forward relevant management enlightenment and suggestions,and elaborates the limitations of this paper in sample object,model construction and disease research. |
