ISYE 6414

Final Project Introduction

Overview

Objective

Perform a professional analysis using everything you've learned in ISYE 6414 and write it up in a professional, academic manner resembling what one would see in a peer-reviewed paper.

Each team member must find and use 3 datasets from different data sources, combine them in an interesting way, and test questions or assumptions about the information they contain for their individual analysis. See the 3d+3d+ requirement for details.

Team Project Requirements

This is a team project — you cannot work solo. You can (and should!) form a team of 5, but if you don't, we'll eventually assign you to one.

Some people will likely drop the course. If groups become much too small, we'll reassign/combine groups to maintain appropriate team sizes.

Everyone is expected to contribute to each element of the project. If you don't contribute adequately, you'll be penalized (up to being given a 0 for that deliverable).

With each of the 3 deliverables, the group will note anyone who does not contribute at least 80% of what is expected. This must be included as a required Appendix to the Analysis Plan and Final Report, and as an additional deliverable in the Code section.

Grading

The Final Project accounts for 35% of your total grade for ISYE 6414.

Grade Breakdown

	Due	Weight
Analysis Plan	Oct 19, 11:59 PM US Eastern Time	30%
Code	Dec 2, 11:59 PM US Eastern Time	20%
Final Report	Dec 2, 11:59 PM US Eastern Time	35%
Peer Ratings	Dec 2, 11:59 PM US Eastern Time	15%

The Peer Ratings are per student, while the Analysis Plan, Final Report, and Code grades will be assigned to each member of a Group.

No Rubrics

Internally, we will use rubrics to grade all of the Final Project components, but we will not release these to deter gaming the rubric and missing out on the joy of structured but fairly open-ended research.

Useful Guidelines

Make sure you pattern your Literature Review on those you see in high quality peer-reviewed research papers (you need 20+ citations). For an excellent example, see Burnam et al. (2014).
Have at least 20 peer-reviewed journal article citations.
Each individual analysis must meet the 3d+3d+ requirement (3 datasets from 3 different data sources, joined together), with at least one dataset having a few hundred rows: part of your job is joining data to search for insights.
You better split at least train & test and report performance metrics. You'd be even wiser if you're going to pick from amongst the various models you'll build as a group by splitting into train, test and validate sets.
We expect you to perform all the professional steps you learn in this class: EDA + outlier screening, GoF testing, variable selection, and attempt to fix data/model when appropriate.
Include graphics when they're useful. Everyone likes pretty pictures: appropriate context and utility is what makes them pretty in data analytical paper.