In this repo, I explore the differences between decision tree regression ML models and random forest regression ML models! The dataset contains information about worldfide FIFA TV viewing audiences and our models' goal is to predict the tv_audience_share of each country!
This is a simple metric to measure the performance of our model. With mean absolute error, we find the average, positive difference between each element's model-predicted value and the actual value. This scheme means that every element's accuracy has equal weight in the final error value.
Each model's MAE will be in the appropriate directory's README.md
I used the random_state parameter when creating the models so everyone running this program will get the same results. random_state ensures that our data breakup (from all the data into training and cross validation segments) is the same no matter when the program is run.
Make sure to have scikit-learn and Pandas installed before attempting to run our program!
Simply run pip3 install sklearn as well as pip3 install pandas to get the necessary modules on your device!
This directory contains the data behind the story How To Break FIFA.
The data file fifa_countries_audience.csv includes the following variables:
| Header | Definition |
|---|---|
country |
FIFA member country |
confederation |
Confederation to which country belongs |
population_share |
Country's share of global population (percentage) |
tv_audience_share |
Country's share of global world cup TV Audience (percentage) |
gdp_weighted_share |
Country's GDP-weighted audience share (percentage) |