A New Method of Technology Foresight on China Engineering Science and Technology to 2035 Based on LDA and t-SNE

Abstract

In the process of technology foresight, the selection of the technology list is completely determined by the expert panel, but without effective tools for experts to understand the division and the overall layout of the technology foresight. Based on the technology list of China Engineering Science and Technology to 2035, this paper designs a new method to support experts to analyze and optimize the technology foresight. First, the LDA (Latent Dirichlet Allocation) topic model is used to extract topics from the technology list of various domains, analyze the main content contained in the technical items of each domain, which could help experts to understand the key point in different subdomain. Then, the “word bag” model is used to generate the vectors of technical items of China Engineering Science and Technology to 2035, and the t-SNE (t-Distributed Stochastic Neighbor Embedding) model is applied for dimension reduction and cluster analysis. Finally, the relationships between the technologies are mapped into 2D image, which could effectively support experts in the overall layout and domain division of the technical items.