Initial Modeling Success

In a previous post, I reported my progress at building an algorithm predicting the probability of success on Kickstarter. I was able to build a model that had 68 percent accuracy on a test set. I also attempted to successfully predict the probability of success with two live Kickstarter campaigns: Love Found and Renegade Repos.

Both campaigns have ended, and our algorithm succeeded in predicting the outcome of both. We gave Love Found an 85 percent probability of success, and it indeed succeeded. In fact, Love Found exceeded its fundraising goal of $3,000 by raising $13,225. This author theorizes that the key to its success was that it was a dance campaign of short duration that set a low goal.

As predicted, Renegade Repo did not succeed. It raised only $475.00. The biggest problem with this web series was that it was a video project of longer duration seeking too much money. Film & video projects are more likely to be unsuccessful, and it did not help that the campaign was over 40 days in duration.