With more than 150,000 crowdfunded projects on Kickstarter, how can you predict if a project will be successful in fundraising? Alex Dixon, Ethen Holzapfel and I – students in EKU’s computer science degree program – set out to answer this very question. Utilizing skillsets developed in the classrooms of EKU, we applied our knowledge of machine learning, artificial intelligence, logistic regression and ingenuity to reach a solution with a computer application.
In CSC 746 – Artificial Intelligence, Alex, Ethen and I used a program called RShiny, a web-based application language for the R coding language, to make an application that predicts the success of crowdfunded projects, or Kickstarter campaign. The application itself allows the user to come up with their own campaign. They can choose the name of their campaign, its category, the country that the campaign is based in, the total amount of money to be raised, and the amount of days that the campaign should run. We used open-source datasets to collect three years’ worth of Kickstarter data from 2016 to 2018. My team and I then used the dataset and linear regression to train the program to predict the success rate of a given crowdfunded project.
My teammates continued to explore more efficient ways to train the program, while I worked on and developed the web based application and its interface. After demonstrating the project in a class for our computer science degree program, Dr. Kim, the CSC 746 professor, asked our team to further develop the application in hopes of releasing it. After updating the application with changes to the user interface, we then presented it at Eastern Kentucky University’s 2019 Computer Science Symposium. While Alex and Ethen gave a poster talk during the symposium, I gave a presentation on our project’s development followed by a live demo of the application.
The application’s development did not stop there. We further trained the program for the application to have higher accuracy in its predictions of success of crowdfunded projects. Additional changes to the application’s user interface were made as well to better allow the viewing of its accuracy and to scale to the screen size of the device utilized.
Our project was entered in the Posters-at-the-Capitol conference in Frankfort, Ky. My team and I were able to display the work produced in the EKU Computer Science Department and demonstrate one of the practical capabilities of machine learning using RShiny to develop web applications like the one that measures success of crowdfunded projects.
The team is still discovering ways to further improve the accuracy rate of our application in hopes of releasing the project officially. View it here.