Start Code 2022
This application was created as a submission for a case presented by Sportradar in StartCode 2022. The case asked for an application that can predict the final score of a football match given data of previous matches.
The application creates predictions through the use of a random forest regression model. The project was created within 48 hours. The app was created with Vue.js as frontend with an API created with FastAPI. The library sklearn was used for creating the machine learning model.
More information about the choice of model and the statistical analysis performed on the trained model can be found in the GitHub attached to the top right icon.
My Role
As I have an interest in learning machine learning and artificial intelligence, I got the responsibility for researching, programming, training, and testing the model. The group divided the work into three sections: data preprocessing, attaining a viable model, and frontend. Therefore, throughout the project, there was a lot of communication and coordination in order to connect each section successfully.
As a group and personally, we were very pleased with the outcome of the project. It was an intense 48-hours; however, we all worked extremely hard to create something that we are proud of. The statistical tests we ran demonstrates that the model was very well suited for the problem.
Training the model
These first two iteration were solely trained on 20% of the dataset.
The third and fourth iterations were trained on 80% of the dataset.
Final Iteration
The final iteration was trained on carefully selected features. Furthermore, similar to iteration three and four, the final training of the model was done on 75% of the training data.
The remaining 25% of the data went to the testing. The results of the statistical tests can be seen in the bottom left.
This project was created in collaboration with Arunan Gnanasekaran arunang2212@gmail.com and Victor Sebastian Immanuel Kremmers Bugge vkbugge@hotmail.com.