[Research purpose]Predicting directed technology fusion relationships based on dynamic knowledge flow characteristics con-tributes to improving the R&D efficiency of enterprises and enables technology management departments to allocate technological resources more effectively.[Research method]Firstly,the first four digits of the International Patent Classification(IPC)codes were utilized as technology units.Directed knowledge flow relationships between technologies were extracted based on the primary class and supplementary class in patents.Subsequently,drawing inspiration from association rule mining algorithms,knowledge flow relationships were filtered based on three dimensions:importance,intensity,and dependence.Knowledge flow relationships meeting the specified threshold criteria were identified as directed technology fusion relationships.Finally,the prediction of directed technology fusion relationships was trans-formed into a supervised binary classification task.Machine learning algorithms served as the modeling foundation,incorporating the knowledge flow characteristics between technologies from the preceding two periods and their temporal variations to forecast the presence of directed technology fusion relationships in the subsequent period.[Research conclusion]The empirical analysis in the additive manufac-turing domain reveals that,employing a cost-sensitive machine learning algorithm for prediction,based on the knowledge flow characteris-tics between technologies from the preceding two periods and their temporal variations,is effective in forecasting directed technology fusion in the subsequent period.As technology advances,the occurrence of one-way technology fusion becomes more frequent,evolving into two-way technology fusion,and even manifesting phenomena of multi-technology fusion.
维普
万方数据