Recommendation of startups as technology cooperation candidates from the perspectives of similarity and potential: A deep learning approach

Companies consistently strive to prepare for new technologies for survival. In a rapidly changing market, absorbing innovation through cooperation strategies can complement internal research and development for new technology development. Startups with state-of-the-art technologies are good candidates for successful cooperation; however, it is difficult to identify their technological positions. Our study suggests a framework to identify appropriate startup candidates using startup profile texts provided by the Crunchbase database. We utilize a doc2vec approach to extract feature vectors representing technological meanings from the startup profile texts and patent abstracts of acquiring companies. Based on these vectors, we apply item-based collaborative filtering to estimate scores for technological similarity between a company and a startup to be acquired. Furthermore, we screen for promising startups using factor analysis, with variables representing the startup's potential. We believe that our framework can save time and effort in the early stage of cooperation planning by supporting effective decision-making.