Research on the state prediction model of crowdfunding projects based on machine learning algorithms

Crowdfunding is a concept that emerged due to difficulties in raising funds for community business projects, social activities, small and micro enterprises, and startups. However, the success or failure of crowdfunding projects is often full of uncertainty. Therefore, predicting whether a crowdfunding project can succeed has become a question worthy of in-depth research. This article takes the crowdfunding website platform Kickstarter as an example and uses machine learning methods to construct an effective crowdfunding project status prediction model. This study aims to compare the performance of different machine learning algorithms in predicting crowdfunding results, while identifying the main factors that affect crowdfunding results and their relative importance. The dataset used in this study includes data from all projects published on Kickstarter between January 2017 and January 2018. This study used six classic classification algorithms to predict the status of crowdfunding activities, and calculated the accuracy of each classification model. The results showed that the accuracy of all six models was close to 1, indicating that they could effectively predict the success or failure of crowdfunding projects. However, the Gaussian Naive Bayes model had slightly lower accuracy than the other five models. Furthermore, the research results indicate that in successful crowdfunding projects, factors such as crowdfunding goals, funds raised, and the number of supporters are more crucial influencing indicators than others.