Abstract: This study develops an algorithmic attribution framework, inspired by biological theory, to explore the effects of various factors on patent maintenance time. It delves into the changing dynamics of these effects across different literature life cycle stages and different technology fields. Employing a combination of survival analysis and interpretable machine learning techniques, this study analyzes a dataset of patents granted in China between 2001 and 2017. Specifically, this study investigates the effects of innate, behavioral and environmental factors on patent maintenance time. The findings shows that the algorithmic attribution framework is effective in accurately determining the relationship between diverse factors and patent maintenance time. Additionally, the study notes that the effects of these factors varies significantly across different literature life cycle stages and among diverse technological domains. The findings can provide critical insights and innovative approaches for patent maintenance time prediction and patent value assessment.

Key words: Patent maintenance time, Algorithmic attribution, Survival analysis, Interpretable machine learning, Patent value