The Design And Implementation Of University-enterprise Cooperation Recommendation System

University-enterprise cooperation is an important way for enterprises to obtain external innovation resources,which is of great significance for enterprises to realize industrial technology upgrading and improve competitiveness.However,in the process of promoting university-enterprise cooperation,it is not easy for technical brokers to match appropriate university experts for enterprises,among which the large number of experts is the main reason for the problems.At present,there are platforms in the market to provide the service of matching experts,but their functions can not fully meet the needs of technical brokers to facilitate university-enterprise cooperation in some scenarios.This thesis designs and implements the university-enterprise cooperation recommendation system to solve the problem of matching between enterprises and university experts when there is a lack of accurate demand text.Based on the patent data of universities and enterprises,this thesis constructs a heterogeneous information network including universities,experts,patents,IPC,engineers and enterprises.Heterogeneous network representation learning method incorporating time function was used to obtain the vector representation of nodes,and the recommendation function was completed by combining the vector similarity of nodes and historical cooperation data.Finally,the Web application is developed based on the Flask framework,and the recommendation results and visual comparison of the two technologies are presented to the technology brokers to help them complete the work of matching experts for enterprises.The main work of this thesis is as follows:(1)Construction and maintenance of heterogeneous information network.Based on the invention patents of universities and enterprises from 2010 to 2020 in the database of CNKI,the author relationship and classification relationship of patents are extracted to construct the network structure.Combined with the time of patent issuance and patent status,the invalid patent nodes are removed and the non-expert nodes of universities are marked to ensure the effectiveness of network data in university-enterprise cooperation application.(2)Designed and implemented the university-enterprise recommendation function.Firstly,in view of the fact that both enterprises and universities have scientific research teams,the community detection method is used to obtain teams of enterprise engineers and university experts,and the team is recommended as a unit.Secondly,considering the timeliness of the patent achievements,the time function was introduced to improve the transfer probability of Metapath2vec++ algorithm in the process of random walk.The node vector representation was extracted based on the feature sequence of nodes,and the recommendation recall was realized by calculating the cosine similarity of the vectors.Finally,the recall results were reordered based on the historical cooperation data to complete the university-enterprise recommendation function.(3)Designed and implemented the university-enterprise cooperation recommendation system.This thesis divides the system into network construction and maintenance module,university-enterprise recommendation module and Web application module.Finally,functional and non-functional tests are carried out to verify the effectiveness of the system.