Skip to content

Latest commit

 

History

History
66 lines (50 loc) · 3.57 KB

final_projects.md

File metadata and controls

66 lines (50 loc) · 3.57 KB
Raw

Final projects

FAQs

  • What's the purpose of the final project? Your final project is an opportunity to apply the techniques we learn during the course to a problem of your choosing. It's also a way to showcase your newly acquired Data Science skills!

  • What types of problems should I tackle? It's best to stick to problems that can be solved using supervised methods, such as predicting numerical values or classes, or forecasting future values of a time series.

  • How much data do I need? To be able to use some of the Machine Learning techniques we cover in the second half of the course, you want 1,000 or more observations. Remember that larger datasets can always be downsampled, but it's difficult to obtain more data when they're not there!

  • Can I use Web scraping? You're free to use Web scraping to collect data; however, be aware that this can be very time-intensive (especially if you've never done it before)!

  • What are some examples of past projects? People have worked on all sorts of projects, for example:

    • Applying Machine Learning techniques to Fantasy Football
    • Characterising spending patterns of customers of an online bank
    • Classifying relevant documents for e-discovery
    • Evaluating regional dependencies of stocks and indices
    • Forecasting supply and demand for an over-the-phone interpreting service
    • Identifying undervalued properties in the London real estate market
    • Modelling cost per click (CPC) data
    • Modelling user activity of a large music streaming service
    • Predicting movie trends and success
    • Predicting stock price movements following press releases
    • Predicting the impact of initiatives against malaria
    • Predicting the likelihood of customer engagement via e-mail
    • Predicting the probability of committing and closing sales opportunities
    • Recommend products to maximise revenue
    • Understanding client churn
    • Understanding how weather affects the online retail industry
    • Understanding whether API requests are made on behalf of Web scrapers
    • Understanding which socioeconomic characteristics help describe higher crime rates in different London boroughs
    • Using pay data from job ads to predict salaries from job titles

  • I have a question not covered here! Get in touch with us! Our feedback mantra is 'early and often'.

Lightning talks

The purpose of the lightning talks is to get some immediate feedback on your project ideas, and to foster collaboration in the classroom.

You should come up with two or three ideas, do some light research (particularly regarding data sources), and prepare some slides covering the following aspects:

  • Problem: What's the problem? Why does it matter?
  • Hypotheses: What do you believe is the solution?
  • Data: What data will you use?

Your lightning talk should last approximately 5 minutes.

Final presentations

The purpose of the final presentations is to share your work with the class, reflect on the experience, and think about how you'd take your project forward.

Your presentation should cover the following aspects:

  • Problem: What was the problem you decided to tackle? Why does it matter to you or your company?
  • Data: What data did you use? Where did they come from?
  • Methods: What kind of analyses did you conduct? Which model(s) worked best?
  • Results: What were the results of your analyses?
  • Lessons learned: What did you learn in the process? What was easy? What was hard?
  • Future work: What are possible extensions of your work?

Your final presentation should last approximately 10 minutes.