Crowdfunding in Indonesia: The Use of Data Mining to Predict Success and Failure (A Case Study)

Abstract

Donation-based crowdfunding is a crowdfunding model that based on social activities and aims to help others. Crowdfunding platform that often used in Indonesia is Kita Bisa. Data shows that every year the number of people who need help by opening donations has increased. However, not all donations are successful and many factors can influence the success of these donations. The purpose of this study is to determine factors that can supporting the successful donations in Kita Bisa, to know how the donations compare before and after the repair, and to determine the performance of the algorithms used in predicting. The method used is classification with C4.5 and K-Nearest Neighbor (KNN) algorithm. Data processing was carried out by experimenting 4 times with a comparison of different training sets and testing sets, and at KNN, experiments were carried out 9 times to determine the best value of k. The results of data processing show that KNN algorithm has a better accuracy, specifically 87.5%. Meanwhile, C4.5 only has 75% accuracy. Both of these accuracy are obtained by comparison of training and test data of 90%: 10% and using a value of k = 5 for the KNN. After obtaining accuracy, the donation is repaired, then compared with the donation before the repair. The result of paired sample t-test results show that there is a difference between the donation before and after repair with the donation after the repair is better. The repaired donation is tested again using the best algorithm (KNN) and predicted to succeed. Thus, it can be concluded that the repaired of donations is able to repair donations that are not successful. Based on the results of data analysis and respondents' opinions, it can also be concluded that the factors supporting the success of donating are quite a lot, with promotions and detailed descriptions being the most important factors.