Technological impact-guided technology opportunity analysis using a generative-predictive machine learning model

Abstract

Although patent analysis-based approaches to technology opportunity analysis have proven useful for discovering unexplored technological ideas, identifying opportunities to maximise the potential of existing technologies remains challenging. This study introduces an analytical framework for identifying new technological domains where existing technologies can exhibit greater technological impact. The proposed framework incorporates a generative-predictive machine learning model integrating the variational autoencoder and multi-layer perceptron architectures. The generative and predictive components of this model are jointly trained to construct an impact-centric technology landscape where technologies with similar domains and impacts are closely located. A gradient ascent search algorithm is used to explore this landscape and identify new technological domains that can maximise the potential technological impact of existing technologies. An empirical analysis covering 133,654 patents related to artificial intelligence technology verifies the reliability and feasibility of the framework in identifying domain-shift opportunities for existing technologies. The proposed analytical framework is expected to serve as a valuable tool that helps firms maximise their existing technological assets in the current era of open innovation.