Kaggle Competition

Outline

Kaggle is a data science community website that hosts numerous guides, datasets, and runs online competitions. Recently they added the feature of Community Competitions where anybody can host a competition using Kaggle's platform.

Kaggle is a fantastic resource and I to encourage you to explore it, this assignment is going to be run using Kaggle's platform.

The assignment is semi-private in that you need a private key to access the competition. This key is listed in the main page on Moodle for this module.

Your task is to analyse the given dataset and construct a suitable classifier. The performance of your classifier is computed based on your uploaded predictions.

Details

First create a Kaggle account (using whatever account name/details you wish) and read the competition instructions (Please ask if there is anything that is unclear).
Go to competition page. This is listed in the main page on Moodle for this module. Accept terms of competition.
Setup a suitable ML project folder.
Download the train and test datasets, train.csv and test.csv and place into subfolder orig.
Split train.csv into your train and test datasets.
Train a classifier using the workflow covered during the course.
Using your model, generate predictions for the data in test.csv and upload.
I will check in with your progress during the semester but the competition runs till end of semester. At end of semester your best submission will be used to generate the final scoring (and your resulting grade). The grading will be based on material you upload to Moodle (see next section).

Submission

Since your Kaggle and Moodle accounts are separate, you will also need to upload to Moodle:

submission.csv your predictions in the format required by Kaggle.
the notebooks/scripts that you used to generate them.

Grading

To allow/reward people for trying reasonable techniques that might not have a positive benefit on the final model score. The grade for this assignment consists of 60% on the model performance and 40% on the workflow in the notebooks uploaded.