Predicting patent transaction cycle using neural hazard model: evidence from technology transactions between companies in South Korea

Technology transactions have become an important strategy for enterprises to adapt to rapid changes in inter-industry technology convergence and market environments. Predicting the timing of technology transactions is essential for enterprises’ technology strategies and for maximizing the value of their intellectual property (IP). However, research on patent transactions timing remains limited owing to data accessibility issues and modeling complexity. In this study, we develop a patent transaction timing prediction model using patent transaction data from the Korea Technology Finance Corporation (KOTEC) and DeepSurv, a neural network-based survival analysis methodology. As a result, DeepSurv shows superior performance (Concordance Index: 0.8946, Integrated Brier Score: 0.0547), which demonstrates its ability to capture complex nonlinear relationships in patent data compared to the Random Survival Forest (RSF) (Concordance Index: 0.8842, Integrated Brier Score: 0.0621) and the Cox Proportional Hazards model (CPH) (Concordance Index: 0.8655, Integrated Brier Score: 0.0836). Additionally, using Shapley Additive Explanations (SHAP), we find that International Patent Classification (IPC), Technology Cycle Time (TCT), and Number of Claims are identified as the key predictive features in predicting deal timing. This study contributes to the existing literature in several ways. (1) This is the first study to apply a neural network-based survival analysis to predict patent transaction cycle. (2) It utilizes real transaction data instead of relying on patent right transfers. (3) It provides actionable insights into optimizing corporate IP strategies and developing policies to activate technology markets. These results can help companies optimize their IP portfolio management, and policymakers should foster more efficient technology markets.