Alex has been taking an online machine learning course lately and I wanted to learn the Wolfram Language and use the cloud platform. So we took the opportunity to take this Kaggle challenge and build a classification model on the Wolfram Cloud.
What it does
- Uploaded ~1000 rows of sample data to the Wolfram DataDrop platform.
- Pulls in the data points and splits them into training (70%) and testing (30%) sets.
- Builds a classification model from the training data using the the Wolfram Classify functionality.
- Runs the classification model on the testing data.
- Calculates the percentage correctly predicted, generally ~70%, and renders a bar chart showing the percentages.
How we built it
- Set up a DataDrop Databin to store the data.
- Parse data with python and post it to the Databin.
- Pulls in the data builds the classification model and verifies correctness all on the Wolfram platform.
Challenges we ran into
- Getting all the data onto Wolfram Cloud
- Learning Mathematica/Wolfram Language
Accomplishments that we're proud of
- Increasing familiarity and intuition for the Wolfram Language.
- Completing first Kaggle challenge.
What we learned
- Wolfram Language - none of us knew it at the start of the weekend.
What's next for Kaggle Data Challenge "SF Crime" with Wolfram Cloud
- Running the algorithm on the official challenge data and submitting the code to kaggle.
Try it out!
Here is the Wolfram notebook file: https://drive.google.com/file/d/0Bz1QgL1Xej6UU2kySnVrbEFxUWM/view?usp=sharing