Research And Application Of Personalized Recommendation Of E-health Service Based On Domain Expertise And Trust

The ever-increasing development of information technology in21stcentury hasfacilitated the appearance of e-health services meanwhile made service consumersinconvenient to find the suitable one himself. Besides, some malicious actors (e.g.,citizens, health professionals or institutions) may publish fake or forged information tomisguide service seekers and make them feel uncertain in decision-making. Thus trustand expertise are two necessary factors for service selection in health area.Traditional collaborative filtering methods do not include trust and expertise aswell as the personal information of users. The similarity metric algorithms only focus onthe rating similarity between items/users, which are too simple and not so accurate.For this purpose, we propose personalized recommendation methods based on ournew user/doctor domain expertise and trust introduced with the features of e-healthindustry. Our research work can be shown as follows:(1)E-health services clustering or health domain forming based on k-mediodsclustering method and current medical ontology.(2)Personalized recommendation methods with new expertise and trust areproposed respectively as well as a combined method with the above two elements.(3)Experiments are conducted to prove the outstanding performance of ourproposed personalized recommendation method using MAE and precision and aprototype system is constructed.The main contribution and innovation can be listed as below:(1)New User/doctor domain expertise degree is defined as the social recognitionand rich experience combined in a certain health area and computed through dataextracted from user-user/user-doctor interaction and usage log.(2)New User/doctor domain trust/trustworthiness is regarded as the passinginformation or decreased risk and computed through new introduced time-aware mutualinformation. Trust propagation in the next step will adopt EigenTrust combined withpersonalized information, such as demographic information and history disease.The result of experiment shows the combined method performs best. Theinfluential factor of expertise is more important in health service recommendationcompared with trust. Besides, the introduction of time difference and demographicinformation makes the final prediction more accurate for trust-based recommendation.