English-Premier-League-Prediction

Apply machine learning to predict English Premier League soccer match.

Demo App

To Run

python3 model.py

Warning: Few python packages required to run the script. Install them all, or use a 3rd party IDE (such as spyder) that pre-installs these packages by default. Few of the packages required are:

• brew cask install chromedriver
• brew install libomp
• pip3 install xgboost
• pip3 install selenium

Scripts

1. clean_data.py

• Includes necessary helper functions to process raw data

2. current_status.py

• Collects and adds more details to the processed raw data
• current/past standings, goals for/against/differences, etc.

3. match_history.py

• Collects the latest match results

4. rankings.py

• Calculate league points and generate standings

5. sofifa_scraper.py

• Scrape overall team stat from FIFA

6. predict.py

• With using processed data, train a ML model to predict future results

7. model.py

• I/O file where the functions from the above files are actually executed

Data

1. data/raw/OVAs (directory)

• scraped overall team stat data

2. data/cleaned/standings (directory)

• historical standing results calculated in rankings.py

3. data/raw/results (directory)

• manually collected historical data of match outcomes
• latest match outcomes of the current season

4. data/cleaned/results (directory)

• data extracted from data/raw/results

5. data/train_data/results (directory)

• data processed from data/cleaned/results

6. data/statistics (directory)

1. data/statistics/round_rankings (directory)
   • standings calculated based on the predicted match outcomes
   • each file in the directory has a date included in its name. It provides predicted standing outcomes at the denoted date
2. data/statistics/prediction_ranking.csv
   • predicted standing at the end of the season
3. data/statistics/prediction_result.csv
   • individual predicted match outcomes
4. data/statistics/round_rankings_summary.csv
   • predicted standing summary over the course of the season

7. data/statistics/best_clf.joblib

• disk cache of classifier that gives the best accuracy of prediction

8. data/database.db

• sql database that stores previous match outcomes, predicted match results and predicted standings

9. data/train_data/final.csv

• csv file used for training a model and making predictions

10. data/statistics/model_confidence.csv

• list of grid searched classifiers and its confidence score