Study And Implementation Of Doctor Recommendation Mechanism Based On Online Searching Of Diseases

Now the Internet has become the main channel for people's getting information.For many pacients,they search the Internet for discovering approprate hospitals and doctors so that they can consult and bespeak before they go to real hospital.In the current network situation,for guaranteeing patients' privacy,it is one of the effective approaches to recommend appropriate hospitals and doctors to patients according to their online disease searching.This thesis studies on the doctor recommendation mechanism based on online searching of diseases,which includes the following work.Firstly,by analyzing the Internet public data,a Web spider is designed to gather vast quantities of information that related to doctors and hospitals.And then the crawled data are reprocessed,which includ purification,renovation,and so on.Furthermore,the disease dictionary is constructed by means of wiping off stop words,Chinese word segmentation,POS tagging,and word frequency ranking.At last the disease dictionary is further improved by manual work.A questionnaire is designed and a survey is conducted for data analyzing and filtrating.Finally,a standard dataset is obtained.Secondly,based on the Gauss-Markov linear model theory,relevant recommendation features are chosen according to the data characters of doctors and hospital.Moreover,the multi-variable linear regression analysis with cross validation is implemented,and the regression coefficient with least square method is estimated.A symbolic model tests with F-test is conducted.The model is predicted and assessed for ensuring the preferential feature and feature weight.Thirdly,based on the doctor data and disease dictionary,a bipartite graph associated network between diseases and doctors is built,and a recommendation algorithm based on the doctor resource allocation and disease resource allocation is proposed.In order to tackle the problem of lack of doctors,two complementary mechanisms are proposed based on the doctor resource allocation and disease resource allocation.Fourthly,based on the feature selection and weight calculation of Gauss-Markov linear model theory,the doctor recommendation results can be predicted effectively.The experimental results also show that the both algorithms can improve the precision of doctor recommendation results.Both of the proposed two bipartite graph based mechanisms can further complement the doctor recommendation result.The proposed recommendation mechanism based on user feedback can also satisfy the users'requirement.Eventually,a prototype system based on above research achievement is designed and implemented.The system combines the modules of patient search,recommendation diagnosis output,detail information of doctors and hospital,recommendation based on patients' further demand,and related disease search.The prototype system can integrate the search functions and provide the results intuitively.Using the system,the patients can efficiently acquire the recommendation results from massive information without privacy leaking.The patients can get great help from the prototype system.