Prediction of Emerging Technologies Based on Analysis of the U.S. Patent Citation Network (1206.3933v3)

Abstract: The network of patents connected by citations is an evolving graph, which provides a representation of the innovation process. A patent citing another implies that the cited patent reflects a piece of previously existing knowledge that the citing patent builds upon. A methodology presented here (i) identifies actual clusters of patents: i.e. technological branches, and (ii) gives predictions about the temporal changes of the structure of the clusters. A predictor, called the {citation vector}, is defined for characterizing technological development to show how a patent cited by other patents belongs to various industrial fields. The clustering technique adopted is able to detect the new emerging recombinations, and predicts emerging new technology clusters. The predictive ability of our new method is illustrated on the example of USPTO subcategory 11, Agriculture, Food, Textiles. A cluster of patents is determined based on citation data up to 1991, which shows significant overlap of the class 442 formed at the beginning of 1997. These new tools of predictive analytics could support policy decision making processes in science and technology, and help formulate recommendations for action.

• The paper introduces a robust citation vector methodology to quantify interdisciplinary patent influence.
• It employs hierarchical clustering with the Ward method to detect emergent technology clusters from the citation network.
• The study validates its approach with a case study that accurately predicted a USPTO reclassification in agriculture and textiles.

Prediction of Emerging Technologies Based on Analysis of the U.S. Patent Citation Network

The paper presents a method leveraging the U.S. patent citation network to forecast emergent technological trends. It introduces a sophisticated methodology that utilizes patents as nodes and citations as edges to form a dynamic and evolving directed network. This network reflects the complex process of innovation by indicating which patents build upon existing technologies, inferred from mutual citations.

At the core of this methodology is the innovative construction of a 'citation vector' for each patent. This vector quantifies how frequently a given patent is cited across various technological domains over time. Each vector component reflects citations from patents in specific technological subcategories, emphasizing non-assortative citations—those from different technological areas. The normalization of these vectors permits the identification of patents with similar technological influence, serving as a measure of similarity for clustering purposes.

The paper employs hierarchical clustering techniques, such as the Ward method, applied to these vectors to detect clusters of patents which may correspond to nascent technology areas. By iteratively evaluating the citation network at different time intervals, the methodology seeks to identify temporal changes in cluster structures indicative of technological evolution.

The practical viability of this approach is demonstrated through a case paper in USPTO subcategory 11 (Agriculture, Food, Textiles). It accurately anticipates the emergence of class 442, validated by a strong correlation between identified clusters and the reclassification undertaken by the USPTO in 1997.

Key elements of the paper include:

• The introduction of a robust citation vector that reflects a patent's interdisciplinary impact, focusing on its non-assortative connections.
• Utilization of hierarchical clustering to identify existing and nascent technological clusters from citation data.
• Empirical validation through backtesting, demonstrating that the model can predict new technological classifications before formal recognition by the USPTO.

These results imply potential integration into policy decision-making, guiding research and development investments by forecasting emergent technological fields. Nonetheless, the paper acknowledges certain limitations, such as the inherent time lag in patent citations and structural simplifications, which may impact prediction accuracy.

Future work could enhance the predictor with improved citation weighting or by applying the methodology across broader datasets. Additionally, exploring the interaction of elementary cluster dynamic events could deepen understanding of the mechanisms driving technological branching. This approach hints at a capacity to uncover guiding principles of technological evolution, potentially applicable to broader complex systems analyses in social and technological domains.