By Sherry Yang and Kevin Velasco
A model that can classify a kickstarter project as likely to succeed or not.
As of April 2019, $4.2bn has been crowdfunded for projects through Kickstarter. The platform has a 36.84% success rate for projects with “games” as the most popular project category.
We are offering a beta model that can produce a likelihood estimation for whether a project will be successful on kickstarter.
- Data Exploration
- Data Cleaning & Encoding
- Create a pipeline
- Create Model 1 using Logistic Regression
- Create Model 2 using Random Forest Classifier
- Compare performance of models relative to our client needs
-
Case 1: A project started in the US with a length of 31 days in the game category with the goal of $20,000 has a probability of succeeding 52% which is better than the overall 35%
-
Case 2: A project started in the US with a length of 29 days in the technology category with the goal of $7,000 has a 26% probability of succeeding compared to the overall 35%
-
Case 3: A project started in Australia with a length of 29 days in the technology category goal of $26,100 has a 10% probability of succeeding compared to the overall 35%