Success Prediction using Random Forest, CatBoost, XGBoost and AdaBoost for Kickstarter Campaigns

Abstract

Kickstarter is a popular crowd-funding platform used by individuals and groups to obtain the seed amount to implement their business ideas. It is essential for the campaign owners to plan their crowd-funding campaign beforehand in order to avoid failure. This paper aims to predict the success or failure of a Kickstarter campaign and help the campaign owners plan better for their campaign by providing trends and analysis based on historical data of campaigns on Kickstarter, ranging from 2014 to February 2019. Various classification and boosting algorithms were applied and it was concluded that Weighted Random Forest along with AdaBoost for the subsampled dataset gives the best accuracy.