Purpose of the Dashboard
The goal of this project is to create an interactive dashboard that allows users to visualize the changing mixed martial art (MMA) styles over the years. On this dashboard, the most important fighting strategies (e.g. head strikes, leg kicks, take-downs, etc.) can be ranked for various years, with the functionality for a user to scroll through how the most important attributes have changed. This can be valuable to fans and historians to understand how the sport has changed over time.
The Data
A dataset was scraped from http://www.ufcstats.com/statistics/events/completedby Rajeev Warrier and was made available on Kaggle (https://www.kaggle.com/rajeevw/ufcdata).The dataset is a list of every UFC fight in the history of the organisation. The original data consisted of each column representing a matchup between two fighters as well as all the fighting statistics (e.g. # of punches) from those fighters. Each matchup row was converted to two rows where each row consisted of a single fighter with their fight statistics and whether or not they won the fight.
Feature Importance
The Boruta package in R was used to determine which fighter’s attributes are the most important in predicting if the fighter wins or loses the fight. Boruta is a feature selection wrapper algorithm which leverages a Random Forest. The method performs a top-down search for relevant features by comparing original attributes’ importance with importance achievable at random, estimated using their permuted copies, and progressively eliminating irrelevant features to stabilise that test.
Findings
From exploring the dashboard, I noted that grappling and wrestling styles of fighting appeared to be more effective in the more recent years, while more flashy striking and kicking dominated the earlier years. This is consistent with the fact that currently six of the seven male UFC champions have an extensive wrestling background.