| The development of personalized medical information service has changed the mode of medical information service. It has been changed from the traditional passive service model of “people looking for information” to “information initiative looking for people” of active service model. Because the information which the system provide to users is based on the customized individual characteristics of users, thus greatly improve the user experience and efficiency of medical information services. However, the existing medical information system is almost based on the traditional Web service, and the construction of user interest model is not combined with the health of the individual user. Therefore, they could not provide prospective medical information service according to health differences of individual users.This paper took advantage of the unique advantages of the mobile terminal equipment, proposes a personalized medical information service system based on mobile Internet. In order to achieve this goal, this paper mainly carried out research work for the following aspects:1. Based on the recommended thinking of user interest model, this paper proposed to construct the user interest model using signs parameter information, and designed the weight rules of the corresponding interest feature item according to clinical standards of the parameters signs, and ultimately realize to provide more comprehensive personalized medical information service for individual users according to their health disparities.2. This paper has completed the construction of user interest model, and proposes two mechanisms of implicit updating user interest model. On one hand, by tracking the user click behavior, using the text feature extraction technology, we realizes the incremental update of user interest model. On the other hand, this paper presents a decay function of interest weight based on the Ebbinghaus Forgetting Curve to achieve the reduction updates of interest model. Through referring to the forgetting law of human brain, the attenuation of users’ interest characteristics item weight, can effectively reduce the weight of the feature item in which users are not interested. It can enhance the accuracy and stability of the expression of interest model, and also improve the adaptive of interest model to avoid the massive and bloated.3. This paper designed a multi-threaded crawler based on WebCollector, which can crawl medical information by classification from the network. By extracting the fingerprint value of the text to remove duplicate documents. The document frequencyinverse document frequency algorithm was used to extract the text feature. And the information of documents was characterized by the vector space model, so as to complete the construction of medical information finally.4. In this paper, with the quartz task scheduling framework provided by Spring, the server can regularly update the user interest model in the background according to the setting time, which greatly improving the real-time of the interest model updating. In addition, in this paper, the recommendation lists of users were stored persistently in the database. Also with the quartz framework for task scheduling, the user recommendation lists were updated regularly, which not only greatly enhance the respond speed of the server during the requests for personalized recommendation, but also enhance the user experience.5. Based on the above theories, this paper designs and implements a prototype system of personalized medical information service based on Android platform. The collection of user interest combined with explicit and implicit ways. Then the vector cosine formula was used to achieve a matching of user interest model calculations with documentation of medical information, and finally achieve the application of personalized recommendation technology in medical information services. System test results show that the precision and recall of information recommendation have reached the requirements of initial use. |
PDF Full Text Request