Projects

Overview
Projects are due by the second lecture of week 7. You should choose one problem from the list of analytics problems and form a small team of
your colleagues to apply computational analytics techniques to solve the problem and produce a summary research report.  The problems are
in order of difficulty:

1) Digit recognition - Which supervised learning classifier is the best hand-written digit detector? [4]
2) Lending Club - Can you build a classifier to automatically reject or accept loan applications? [6]
3) Zillow - Can you find undervalued houses in a neighborhood? [8]
4) D4D  - What rules and relationships can you find in call data from the Ivory Coast? [10]

(key:
[10] = expert, [1]  = beginner)

Problem Solving

Each problem requires first extracting data from a specified text-based or web repository and cleansing and transforming it into R readable format. Each team shall follow the work-flow guidelines for applying analytics as set out the lectures. The team will need to answer a question or hypothesis, describe the data and apply classification techniques to arrive at the answer. A report will address, but is not limited to, the following checklist of
guideline questions:

- How would you describe the available data?
- What is the problem that you are solving?
- Can the problem that you are solving be directly solved or did you need to consider another problem first?
- What, if any, was the critical step in solving this problem?
- What are the assumptions and limitations of your approach and what makes it better to any other approach?
- What conclusions can you draw from your work and how might you build on your results if you had more time?

Group Work
You are strongly advised to closely co-ordinate with your colleagues throughout the project and distribute the work load fairly and commensurate with each team members' interests, experience and background.  Budget for twice the amount of time that you think a task will reasonably take to complete and be pre-emptive about any potential difficulties in synchronizing your work.

Grading
Because project grades are assigned to the team and not on an individual basis, it is vital that each team member is accountable for their portion of the project and does not let down their team. Graded projects will be returned by week 8.

Final Report Checklist
1) Names of all project members
2) Concise executive summary highlighting the main results and conclusion
3) Description of your methodology and dataset
4) Description of your final results and the code used to produce the results
5) Summary and conclusion
6) Appendix with intermediate results and code samples
Comments