Research On Technology Opportunity Identification Method And Application Based On Link Prediction ——Taking Privacy Protection Technology As An Example

In the era of knowledge economy,scientific and technological strength,as an important part of national core competitiveness,determines a country's position in the international pattern.It is a basic and important requirement for China to improve its technological sensitivity and quickly and accurately identify the direction of scientific and technological development.The identification and mining of technological opportunities have laid an important foundation for the state to formulate guidelines and policies for scientific and technological R ? D,support universities and enterprises to carry out R ? D,and also put forward higher requirements for subjects providing scientific and technological information and knowledge supply.In previous studies,the methods used in technology opportunity identification are mostly from a macro point of view to draw the general direction of technology opportunity development,lacking specific guidance and detailed data support.As one of the important topics of graph mining,link prediction method can quantify whether there will be links between nodes and the strength of link relationship.Link prediction between IPC nodes based on international patent classification number can provide specific and detailed data support for technology discovery.Compared with the macro development direction suggestions in the research field of technology discovery methods,the link prediction based on IPC co-occurrence can give the future development direction of IPC portfolio from a more micro perspective,and can become an important auxiliary and supplement to the macro suggestions.Therefore,in order to timely and effectively identify technology opportunities and provide scientific and technological information support for the layout of technology R ? D resources,this paper selects the privacy security topic that has attracted much attention in the era of artificial intelligence,and carries out the research on technology discovery methods based on the relevant patent data in the field of privacy protection.This study is mainly divided into the following two parts:The first part is the experiment of link prediction based on GNN.GNN can bring the characteristics and path information of nodes into training at the same time,and has strong ability to represent nodes and links.In order to improve the training speed and generalization performance,this paper selects the sampled graph convolution network Graph SAGE,and takes the three characteristics of IPC time dimension: IPC cooccurrence annual distribution,IPC time attention and IPC time succession as node characteristics to improve the node presentation ability,and customize the link prediction score function to predict whether IPC can be used as a combination of technical elements in the future.After determining the relatively optimal experimental parameters,it is compared with five methods based on node similarity: CN index,RA index,Jaccard similarity coefficient,Adamic-ADAR index and PA index,as well as graph walk algorithm node2 vec,which confirms the Graph SAGE?T in this paper,and gives suggestions on the relevant parameters affecting the model and application scenarios.The second part selects the core IPC number in the field of privacy protection for the identification,application and analysis of technical opportunities according to the results of link prediction.It mainly adopts the SNA method and statistics method.Firstly,it makes an overall analysis on the relevant technologies in the field of privacy protection,and selects the core IPC as the object to analyze the link prediction results through the analysis centrality.Through calculation and visualization,the change of link strength in the future is confirmed.The technological opportunities for sustainable development,technological opportunities for breakthrough and innovation and technological opportunities for transformation are identified from the perspectives of predicting the future maintenance of original links,the generation of new links in the future and the extinction of old links in the future.The occurrence frequency of technology combination elements in recent years,in different economies.The application of industry and other scenarios,as well as the opportunities or challenges faced in the selection of materials and processes,the improvement of algorithms,and the innovation of different types of application scenarios in the future,give some suggestions for the technical layout and the allocation and scheduling of technical resources.This study verifies the effectiveness of the graph neural network model considering time characteristics in the link prediction method on the patent data set,and also gives a new idea for the identification and analysis of technology discovery based on the link prediction results.Taking the field of privacy protection as an example,this study identifies technology opportunities from a more micro perspective,a more fine-grained level and a more quantitative result,so as to provide assistance for tasks such as technology discovery and technology foresight,It can provide reference for corresponding technology application,R ? D and innovation of University enterprises.