In this project, you will pick a topic of your choosing and perform an end-to-end analysis using what you have learned. You will apply the statistical and machine learning techniques we have learned over the last few weeks and present your results to all of us.
- Ask interesting and thoughful questions and find the data to answer them.
- Try to find multiple source for your data to make a more complete analysis.
- Focus on improving in areas that are hard for you or learning more about something with which you feel comfortable.
- Apply the statistical and machine learning techniques we have learned.
- Create useful and clear graphs.
- Present your insights in a thoughtful, clear and accurate way.
- You must plan your project. That is why creating a Trello Board is mandatory.
- You CAN'T CODE until you project is planned.
- This is an individual project.
- It is strongly suggested that you have a rigorous analysis.
- A well-commented notebook with your analysis (Jupyter or Kaggle).
- A 5 minute presentation in the classroom (+2 minutes of questions).
- Repository with your workflow + documentation + code. Even if you are working alone, you need to keep good practices!
- The database where you have kept your data.
Wednesday - Thursday
- Think about a topic and propose some questions.
- Choose data that is relevant to your questions.
- Look for documentation to give context to your project.
- Write the README file in your repository.
- Get approval for your project
- DO NOT START CODING
NO CODE UNTIL HERE
Friday - Wednesday
- Start importing the data and cleaning it.
- Start the analysis. Remember all the techniques you have learned!
- Prepare a draft of your first slides presentation (no analysis or conclusions yet): title, motivation, context, ... It is good practice to add the results of your analysis as soon as you obtain them.
Thursday
- Rehearsal. Take the feedback and use it!
- Finish the analysis. Finish the slides.
- Final improvements!
Friday
- Presentation!
- 5 minutes presentation in the classroom (+2 minutes of questions).
- Keep It Simple!!!!
- Organize yourself (don't get lost!). Respect deadlines.
- Ask for help vs Google is your friend.
- Define a simple approach first. You never know how the data can betray you ๐.
- Document yourself. Learn about the problem and what research has been done before you.
- Before making a graph, think what you want to represent.
- Don't force yourself to use tecniques if they are not helpful for your objective.
- If using machine learning, remember:
- This is an iterative process. Try your best to improve your model performance by:
- Try different models and select one that is the simplest yet produce the best result.
- Try different hyperparameters and see if they improve the result.
- This is an iterative process. Try your best to improve your model performance by: