Medical community expert classification based on potential semantic feature transfer learning

Nowadays, the mobile medical community, providing a communication platform for medical, medical treatment, pharmacy, life science as well as other related domains, acts as a professional social network for doctors, medical institutions, healthcare practitioners and life science. In the medical community, users can ask questions and receive the response from a professional doctor. It is possible to push the user's question to the specific doctor via classifying medical experts in the community, Therefore, the user's question can be answered in time. In the training of classification model of medical experts, the general approach is supervised learning. However, several different domains, including respiratory and chest diseases, first aid and critical illness and neuroscience, could be found in the medical community. Classical supervised learning algorithms find good classifiers for a given learning task using labeled input-output pair and require a large number of labeled training samples, when labeled data is limited and expensive to obtain. However, the original classification model can not obtain the optimal effects n the new domain. Moreover, trace number of categories or unclassified need to be re-labelled, leading to a higher price. In order to address the problem of cross-domain expert classification model in medical community, we combine user's inherent information as the keyword together with user's potential information, thereby improving the cross-domain model of medical community expert classification. Through the data collected in the medical community, our experimental results suggest that the method can be used to achieve better classification effect in small or unlabeled new domains.