Prediction of Crowdfunding Project Success with Deep Learning

Abstract

Over the past century there has been a dramatic increase in crowdfunding activity, which offers an alternative for both creators and backers to sell products and invest in creative businesses respectively. However, empirical analysis shows that only one-third of crowdfunding campaigns could meet their fundraising goal. The aim of this paper is to develop a model that predicts the success of crowdfunding project with deep learning. The datasets are retrospectively collected from Kaggle and contain historical records of Kickstarter campaigns. The model could provide insights in pre-lunching stage and in early stage of fundraising. The proposed MLP model can provide accountable results when applied to different crowdfunding platforms that have not been addressed before. Comprehensive experiments are conducted and a variety of classification algorithms have been tested to support this prediction engine and they concluded that the MLP model has the best outcome with the highest degree of confidence.