PREDICTING TECHNOLOGY SUCCESS BASED ON PATENT DATA, USING A WIDE AND DEEP NEURAL NETWORK AND A RECURRENT NEURAL NETWORK

Marie Saade (1) , Maroun Jneid , Imad Saleh (1)

1 PARAGRAPHE - Laboratoire Paragraphe

The temporal dynamic growth of technology patents for a time sequence is a major indicator to measure the technology power and relevance in innovative technology-based product/service development. A new method for predicting success of innovative technology is proposed based on patent data and using Neural Networks models. Technology patents data are extracted from the United States Patent and Trademark Office (USPTO) and used to predict the future patent growth of two candidate technologies: "Cloud/Client computing" and "Autonomous Vehicles". This approach is implemented using two Neural Networks models for accuracy comparison: a Wide and Deep Neural Network (WDNN) and a Recurrent Neural Network (RNN). As a result, RNN achieves a better performance and accuracy and outperforms the WDNN in the context of small datasets.

Mots clés: Innovative product development; forecasting technology success; patent analysis; machine learning

Domaines: Sciences de l'Homme et Société

Marie Saade, Maroun Jneid, Imad Saleh. PREDICTING TECHNOLOGY SUCCESS BASED ON PATENT DATA, USING A WIDE AND DEEP NEURAL NETWORK AND A RECURRENT NEURAL NETWORK. IBIMA 33, 2019, Granada, Spain. ⟨hal-03004809⟩