AI-Driven Patent Value Prediction: Beyond Citations to Contextual and Strategic Signals

Li et al. (2023) identify low-value patents as a source of inefficiency, but most value prediction models rely heavily on citation counts. Inspired by recent advances in AI for patent analytics (Podrecca et al., 2024), this research would create machine learning models incorporating richer features: litigation history, licensing deals, market entry timing, regulatory changes, and even social media or news sentiment around a patent’s domain. By training and validating these models on large-scale, multi-country datasets, the research could significantly outperform citation-based approaches. The innovation here is methodological—expanding the feature set and context for value prediction—offering firms, VCs, and policymakers a much sharper tool for allocating resources to high-impact innovation. This could transform how innovation portfolios are managed and optimized globally.

References:

1. Identify more inefficiency sources to firm's innovation performance: low-value patent or R&D expenditure congestion?. Zhifu Li, Hao Zhang, Anyu Yu, Ran Liu (2023). Technology Analysis & Strategic Management.
2. Artificial Intelligence for Climate Change: A Patent Analysis in the Manufacturing Sector. Matteo Podrecca, G. Culot, Sam Tavassoli, G. Orzes (2024). IEEE transactions on engineering management.