A cosine similarity based expert recommendation method using a hybrid collaborative filtering technique which calculates the similarity between an expert's attributes (i.e. specialties and field of interest) and the experts and recommends those experts related to a user is disclosed. A high-precision image matching device and a method thereof using skin markers and inbody feature points, as an example, is capable of performing an expert recommendation process based on a score estimation method when the volume of data stored in an expert database is insufficient, improving the quality of recommendation by applying weight based on the activities (i.e. papers, career, etc.) of an expert in order to determine the expert's expertise, and allowing an SNS user to communicate and work with others regarding his/her field of interest by getting a recommendation about the experts regarding his/her field of interest. [Reference numerals] (AA) User information;(BB) Recommendation score table;(CC) Score estimation table;(DD) Ranking;(EE) Verification;(FF) Expert recommendation
展开▼