The capstone is an opportunity to demonstrate your ability to perform and present the knowledge and growth of twelve weeks of a very intense class. Building a neural network model for something obscure and cool may be great, but choosing something that can exercise a broader range of skills is strongly recommended.
- Data collection
- Data munging
- EDA
- Feature engineering
- Modeling / machine learning
- Model evaluation
- Interpretation
- Visualizing and communicating results
- Be prepared to discuss why the models you chose make sense, and how the data work with it given the goal.
- Define your problem statement.
- After articulating your problem statement, outline your goals and success criteria.
- Describe 1-2 potential datasets that address your problem statement. Identify the source and the format of your dataset(s).
- Identify a potential audience of stakeholders who may be interested in your findings.
- Solve your problem!
- Create a 12-15 minute presentation slide deck. This slide deck should be accessible to a wide audience - especially since you'll likely be the only subject-matter expert in the room. However, you'll also want to include details so that we understand your thought process and how well you were able to solve your problem.
- Be prepared to discuss and defend your work... from your choice of dataset to your model-building decisions to your conclusions. They're all fair game!
- Include your slides in your portfolio.
- Create at least one blog post about your findings.
The data for your project is the single most difficult problem you will face. The potential to tell a story, build predictive models, or even brainstorm, will be dependent on the actual dataset that you will use.
You may not know what you can do until you get a good set of data. It's a good idea to look around for datasets as early as possible. Completing your capstone will be entirely contingent upon this data - your data sets the maximum and minimum for what you can achieve.