Prediction of fundraising outcomes for crowdfunding projects based on deep learning: a multimodel comparative study

Abstract

As a new financing model, crowdfunding has been developed rapidly in recent years and has attracted the attention of investors and small- and medium-sized enterprises and entrepreneurs. However, many projects fail to be funded; thus, crowdfunding project fundraising outcomes forecasting and multimodel comparisons are meaningful ways to identify project quality and reduce market risk. It is important to reduce participation risk through automated methods, which is of great significance to the sustainable development of Internet finance. First, based on the data from the Kickstarter, preprocessing and exploratory analysis are conducted. Then, we introduce a deep learning algorithm (multilayer perceptron) and apply it to the prediction of crowdfunding financing performance. We compare deep learning with other commonly used machine learning algorithms, including decision tree, random forest, logistic regression, support vector machine, and K-nearest neighbors algorithm. We tune each machine learning algorithm to get the best parameters. The experimental results show that the deep learning model can obtain the best prediction results, with an accuracy of 92.3% when predicting the fundraising outcomes of crowdfunding financing, followed by the decision tree. Deep learning shows significant advantages in many evaluation criteria, which demonstrates the potential for crowdfunding project financing predictions. This study combines machine learning with Internet finance, providing inspiration for future research and resulting in many practical implications.