Innovative Reward-Based Crowdfunding Decision Model

Abstract

Due to the risky nature of newly creative projects for entrepreneurs, reward-based crowdfunding is currently an alternative fundraising channel for those who need seed funding to finance the creation of their prototype. The objectives of this research are to explore the success factors, including entrepreneurial, project and campaign factors, in project fundraising under a reward-based crowdfunding platform. We propose to develop a model for predicting the success of crowdfunding projects by machine learning. The datasets have been retrospectively gathered from historical records of campaigns in the Kickstarter website. The study's findings show that the logistic regression and decision tree models, respectively, had accuracy rates of 88.2% and 88.8%. The highest accuracy percentage of 94.1% originates from new testing data that has been externally validated for the technology industry. The practical implication of our research is that entrepreneurs can apply the proposed prediction model to identify the most influential topical features embedded in campaigns.