Machine Learning for Creativity: How Similarity Networks Help to Design Winning Projects in Crowdfunding?

Abstract

While crowdfunding has grown enormously, creating successful projects remains a major challenge. Few studies have shown how to improve a project proposal when designing a project. One potential yet unexplored approach is to apply the concept of combinatorial creativity, analyzing a new project in connection to prior similar projects. Applying machine learning techniques (Word2vec and Word Mover’s Distance), we measure the degrees of similarity between projects on Kickstarter. We analyze how this similarity pattern relates to a project's funding performance. We find: (i) the prior level of success of similar projects strongly predicts a new project's funding performance, (ii) the funding performance increases with a balance between being novel and imitative, (iii) the optimal level for funding goal is slightly above the funds raised by prior similar projects, and (iv) the funding performance increases with a balance between atypical and conventional imitation. We implement these findings to generate actionable insights for project creators and crowdfunding platforms.

Keywords: crowdfunding, similarity, networks, Word2vec, funding goal, imitation