Forecasting the Emergence of a Dominant Design by Classifying Product and Process Patents Using Machine Learning and Text Mining

Forecasting the emergence of a dominant design in advance is important because the emergence of the dominant design can provide useful information about the external environment for the product launch. Although the emergence of the dominant design can only be determined as a result of the introduction of the product into the market, it may be possible to predict the emergence of the dominant design in advance by applying a solution based on patent analysis. In the newly proposed technique of separating patents, we can capture changes in the state of technological innovation and analyze the emergence of the dominant design, but there is a problem that it requires processing of large amounts of patent data, and that the processing involves subjective judgments by experts. This study focuses on analyzing technological innovation trends using an approach that separates product patents from process patents, investigates whether this approach can be applied to machine learning, and aims to develop a learning model that automatically classifies patents. We applied text mining to patent information to create structured data sets and compared nine different machine learning classification algorithms with and without dimensionality reduction. The approach was effectively applied to machine learning, and the Random Forest, AdaBoost and Support Vector Machine models achieved high classification performance of over 95%. By developing these learning models, it is possible to objectively forecast the emergence of a dominant design with high accuracy.

@article{Masuda2025,
  title     = {Forecasting the Emergence of a Dominant Design by Classifying Product and Process Patents Using Machine Learning and Text Mining},
  journal   = {International Journal of Advanced Computer Science and Applications},
  volume    = {16},
  number    = {1},
  year      = {2025},
  publisher = {The Science and Information Organization},
  author    = {Koji Masuda and Yoshinori Hayashi and Shigeyuki Haruyama},
  doi       = {10.14569/IJACSA.2025.0160104},
  url       = {https://doi.org/10.14569/IJACSA.2025.0160104}
}