Week 4 Presentation
Patrick
January 3 2017
Everyone Is Awkward
The world has undergone some incredible social changes over the past 30 years. First, the majority of the population has moved away from small towns, breaking all childhood social bonds, and concentrated in big cities. Also, people have largely abandoned live interpersonal interraction for social media.
As a result, dating is broken. The world's cities are comprised of millions of people… all smushed together… and everyone is alone. Young people have forgotten how to date.
The Perfect First Data
This application is meant to be a fun game that can serve as a guided first date.
We supply 29 personal and demographic questions that serve as great conversation starters.
If there's any chemistry between you and whatever person/machine/animal you've decided to date, we use machine learning to calculate the ideal next date based on your partner's predicted preferences.
Click Here To Try It Out
How It Works
We've taken survey data from the 2013 FSEV UK survey. This extensive survey contained over 150 questions. We've narrowed these questions down to just 29 that could be used to determine a person's musical tastes using a Random Forests model.
From this survey daa, the same questions could also be used to predict the subject's favorite movie genres, pet peeves, personal interests and more. Although this feature has been left out of the application for now.
If you know your date's favorite movie genres and music styles, you can use this information to plan the perfect second date.
Click Here To Try It Out
Example Prediction
For every style of music represented, we've created a model. These models are saved to files, and saved these files for the live web application. In order to use them, we load all the models and run each through the input data.
Here is an example, based on one of the training data examples from the data.
load("musicCountryFit.RData")
predict(musicCountryFit, newdata=musicCountryFit$trainingData[19,])
19
2.388237
According to this prediction this particular user would rank Country Music with a score of 2.4 out of 5.
(We've cached this result because the source files would've been difficult to run directly from other machines. But this is an accurate representation of the application's inner workings)