A new CSIL working paper authored by Emanuela Sirtori shows how community detection analysis of a network of patent technology codes (IPC) can be used to classify a large and sparse patent database in a fully automated and unsupervised way. The research aligns with the increasing demand for efficient methods to identify meaningful technological domains that can inform the study of technologies and their evolution.
Light Emitting Diode (LED) has been selected as a test-bed of the methodology. As a multi-purpose technology, LED evolved for decades, spanning several technology fields, before finding mass application in the general lighting industry. The analysis has been conducted over the largest database of patents related to LED ever used in the literature, covering over 400 thousand patent documents filed in 77 patent offices in the world between 1962 and 2018. VOS and Louvain community detection algorithms have been applied to find the technology domains around which the patent activity concentrated across the long and multi-directional historical evolution of LED.
Results have been compared with other studies and approaches in order to highlight the advantages of the proposed methodology. IPC-based community detection proves particularly useful in classifying other
technologies characterised by a meandering evolutionary process across several domains. It does not require particularly advanced data science skills and allows the flexibility of choosing the level of granularity in the classification by adjusting the resolution parameter.
Sirtori E (2023), “Identification of multiple technology domains of LED through IPC community analysis”, Working Papers 2023/02, CSIL.