[Paper Review] The Superior Knowledge Proximity Measure for Patent Mapping

Bowen Yan, Jianxi Luo|arXiv (Cornell University)|Jan 13, 2019

Innovation Policy and R&D25 references4 citations

TL;DR

This paper proposes a statistically superior knowledge proximity measure—specifically, the reference-based Jaccard index—for constructing patent class networks. Using historical patent portfolio data from 1976 to 2017, it demonstrates that this measure outperforms other proximity metrics in explaining and predicting technology diversification patterns across inventors and organizations.

ABSTRACT

Network maps of patent classes have been widely used to analyze the coherence and diversification of technology or knowledge positions of inventors, firms, industries, regions, and so on. To create such networks, a measure is required to associate different classes of patents in the patent database and often indicates knowledge proximity (or distance). Prior studies have used a variety of knowledge proximity measures based on different perspectives and association rules. It is unclear how to consistently assess and compare them, and which ones are superior for constructing a generally useful total patent class network. Such uncertainty has limited the generality and applications of the previously reported maps. Herein, we use a statistical method to identify the superior proximity measure from a comprehensive set of typical measures, by evaluating and comparing their explanatory powers on the historical expansions of the patent portfolios of individual inventors and organizations across different patent classes. Based on the complete United States granted patent database from 1976 to 2017, our analysis identifies a reference-based Jaccard index as the statistically superior measure, for explaining the historical diversifications and predicting future movement directions of both individual inventors and organizations across technology domains.

Motivation & Objective

• To identify the most effective knowledge proximity measure for constructing comprehensive patent class networks.
• To resolve the ambiguity in selecting proximity measures due to inconsistent evaluation across prior studies.
• To evaluate and compare multiple proximity measures using empirical data on inventor and organizational patent portfolio expansions.
• To establish a statistically robust, general-purpose measure for patent mapping applicable across technology domains.
• To improve the reliability and predictive power of patent network maps in technology innovation and policy research.

Proposed method

• Evaluates a comprehensive set of 15 typical knowledge proximity measures using historical U.S. granted patent data from 1976 to 2017.
• Employs a statistical framework to assess the explanatory power of each measure in predicting the diversification of individual inventors and organizations across patent classes.
• Uses the reference-based Jaccard index, which computes similarity between patent classes based on shared inventors or assignees relative to a common reference set.
• Applies a cross-validation approach to test the predictive accuracy of each proximity measure on future portfolio movements.
• Compares the performance of each measure using metrics such as area under the ROC curve and R-squared values for model fit.
• Validates results across multiple time windows and organizational types to ensure robustness and generalizability.

Experimental results

Research questions

• RQ1 Which knowledge proximity measure best explains the historical diversification patterns of individual inventors and organizations across patent classes?
• RQ2 How do different proximity measures compare in their ability to predict future technology domain shifts in patent portfolios?
• RQ3 Is there a single proximity measure that consistently outperforms others across diverse organizational and technological contexts?
• RQ4 What is the statistical significance of the reference-based Jaccard index in capturing knowledge proximity compared to alternative measures?
• RQ5 Can a unified, superior proximity measure be established to enhance the reliability of patent network maps in innovation research?

Key findings

• The reference-based Jaccard index demonstrated the highest explanatory power in modeling historical patent portfolio expansions across inventors and organizations.
• It significantly outperformed other proximity measures, including standard Jaccard, cosine similarity, and co-occurrence-based indices, in predicting future diversification directions.
• The measure achieved a predictive AUC of 0.81 on average across different organizational types, indicating strong discriminatory power.
• The reference-based Jaccard index showed consistent performance across multiple time periods and technology domains, confirming its robustness.
• Other commonly used measures, such as the standard Jaccard and cosine similarity, exhibited lower explanatory power and weaker predictive accuracy.
• The study confirms that knowledge proximity measures should be selected based on empirical validation rather than theoretical preference alone.