This repository contains the code and resources for a machine learning project that aims to predict the chance of a football team winning a match.
The Football Winning Percentage Prediction project utilizes machine learning techniques to forecast the likelihood of a football team winning a match. The project leverages historical data from the past three seasons (2020-21, 2021-22, and 2022-23) to train and evaluate the prediction models.
The dataset used for training the machine learning models was obtained by scraping data from FBref.com. The scraping process is documented in the scraping.ipynb file, which provides details about how the data was collected using the Beautiful Soup library.
The prediction models are implemented in the prediction_model.ipynb notebook. Three different models were trained and evaluated for this project: Random Forest, XGBoost, and Logistic Regression. The notebook contains the code for each model, including data preprocessing, feature selection, and model training.
To provide easy accessibility and user interaction, a web application was created using the Streamlit library. The application is implemented in the app.py file. Users can input the relevant features, and the trained models will predict the winning percentage of a football team in a match.
To access the web application, follow these steps:
- Clone this repository to your local machine using the following command:
git clone https://github.com/2spi/pl-predictor.git - Install the required dependencies by running the following command:
pip install -r requirements.txt - Run the web application using the following command:
streamlit run app.py