Success Prediction of a Crowdfunding Project in Art Categories

Abstract

Crowdfunding is a trending topic that expresses an alternative financial source for entrepreneurs. There are several crowdfunding platforms worldwide and a huge number of projects targeting to collect the required amount of money for starting up innovative ideas. This paper is expected to guide project owners who makes online calls to collect funding. It serves this purpose by identifying explanatory attributes that brings success to a crowdfunding project, predicting outcomes of projects and constructing decision rules for identifying successful project outcomes. The study focuses on reward-based crowdfunding on art, comics, dance, film & video, music and theater categories. The dataset used in the analysis contains 8996 projects and 19 attributes. Data management and analysis were performed using Jamovi and WEKA. We used feature selection to determine important attributes. In line with the results derived from feature selection; videos, updates, comments, rewards, goals, number of images, number of word attributes has been included in the analysis. Applying decision trees, the results of the analyses comply with previous studies’ findings on the positive effect of some attributes –having at least one video, number of comments and updates. Our model works with 74.5554% accuracy. In comparison to the previous studies, experimental results show satisfactory accuracy and correct classification rates.