Unveiling the interaction landscape and innovation-driven relationships in the support vector machines field from a knowledge network perspective

Purpose This study explores the evolving interactions between scientific research and technological innovation in the support vector machines (SVMs) field, aiming to reveal their coupling mechanisms and innovation-driven dynamics. Design/methodology/approach We analysed 84,369 publications and 14,271 patents. Using rapid automatic keyword 123,514 keywords were extracted (109,996 scientific, 13,518 technical and 2,199 shared) to construct knowledge networks. Coupling strength, structural entropy and effective transfer entropy (ETE) were applied to assess network integration and directional influence between science and technology (S&T). Findings The scientific network is consistently larger yet sparser than the technical one (average density 1.08E−03 vs 1.19E−02), though both show small-world effects (clustering coefficients: 0.796 vs 0.778). Coupling has intensified over time, reaching 0.5116 between the 2023 scientific and 2022 technical networks. ETE results indicate two phases: from 2004 to 2017, science strongly drove technology (ETE = 0.2888), while from 2018 to 2023, technology exerted a weaker reverse influence on science (ETE = 0.0153). Structural entropy analysis highlights stable scientific communities but volatile technical ones, especially after 2017. Originality/value By integrating content and structural perspectives, this study quantitatively demonstrates the shift from science-driven to technology-driven innovation in SVMs, offering new insights for research management and policy formulation.