Backer Preference Modeling and Prediction of Crowdfunding Campaign Success

Crowdfunding market has witnessed rapid growth; however, the overall success rate remains relatively low. Therefore, predicting crowdfunding success has garnered significant attention, offering benefits for backers, platforms, and campaign creators in terms of risk reduction and optimized resource allocation. Departing from the traditional campaign-centric perspective, we proposed a novel backer-centric approach for predicting crowdfunding success: Modeling backer preference based on their pledge history and aggregating preference of active backers within categories for predicting crowdfunding success. To validate the performance of this approach, we identified backers who participated in the discussion of recently popular campaigns on Indiegogo platform. A total of 7895 backers were captured along with the 91299 campaigns they either followed or contributed to. Additionally, we gathered 73581 campaigns for training and testing predictive models. Results indicate that considering backer preferences significantly enhances that accuracy of crowdfunding success prediction models. Further, we observed that modeling backer preferences based on higher activity level can further enhance the performance of the predictive model.