Quantifying Learning and Competition among Crowdfunding Projects: Metrics and a Predictive Model

Abstract

The performance of a crowdfunding project is highly situational-dependent. In this study, we quantify the interactions between crowdfunding projects in order to understand how these interactions can help predict the performance of crowdfunding campaigns. Specifically, we utilize Natural Language Processing (NLP) techniques to create a semi-automated system to label the associated product for each crowdfunding campaign. We also propose three sets of metrics to measure how crowdfunding projects learn from and compete with each other. Finally, we propose a machine learning model and demonstrate that the proposed metrics and the proposed model outperform other combinations when predicting the performance of crowdfunding projects.