This analysis is my first Capstone project in the Springboard Data Science Career Track course. The primary dataset for this analysis is NFL Combine (Combine) data. The Combine is an invitation-only showcase for aspiring NFL players. They perform various physical and mental drills and tasks. The Combine is a part of the total evaluation process by NFL teams prior to the seven-round NFL Draft. This completed project uses these Combine results to predict which round a player might be selected in the NFL Draft.
This repo contains the code, data, documentation and reports for this project.
Using a metric of F1 score (in this case with a weighted average), various models are tested to predict the multi-class output. A stacked ensemble classifier is found to perform best. It achieves a score of 0.29, improving upon the baseline score of 0.19.
| Step | Description | File(s) |
|---|---|---|
| Proposal | Full project proposal and idea | report |
| Data wrangling | Data cleaning, wrangling, and munging | code, report |
| Data story | Looking into Quarterback trends | code |
| Inferential statistics | Closer statistical look between various distributions | code, report |
| Milestone report | Summary of all steps completed so far | report |
| Model building | In-depth analysis and machine learning | code |
| Final report | Discussion and overview of completed project | report, slide deck |
| Bonus Prediction | Predicting 2018 draft with final model | code |