PREDICTING FOOTBALL MATCH RESULT OF ENGLAND PREMIER LEAGUE WITH ARTIFICIAL NEURAL NETWORK
Football (Soccer) is a well-known sport across the globe. Almost every country has its own league. Generally, there is no certain way to predict the outcome of a football match. This research covers predicting football match results with mathematical methods, Artificial Neural Networks. Artificial N...
Saved in:
| Main Author: | |
|---|---|
| Format: | Final Project |
| Language: | Indonesia |
| Online Access: | https://digilib.itb.ac.id/gdl/view/47815 |
| Tags: |
Add Tag
No Tags, Be the first to tag this record!
|
| Institution: | Institut Teknologi Bandung |
| Language: | Indonesia |
| Summary: | Football (Soccer) is a well-known sport across the globe. Almost every country has its own league. Generally, there is no certain way to predict the outcome of a football match. This research covers predicting football match results with mathematical methods, Artificial Neural Networks. Artificial Neural Network is a method of Deep Learning which able to find the best model to fit almost every problem. This research is a continuation of prior researches which studied this topic with other mathematical methods. The objective is to learn how to architect artificial neural network model to predict English Premier League matches, then identifying features or variables that affect the model which can be used for further examination. The scope of this research is only English Premier League football matches throughout the 2008/2009 season to 2019/2020 season. The data is taken from www.football-data.co.uk which provides football match data records from all leagues in almost every country. Feature engineering is done to the match data records before inputting it to the model, transforming into features which is calculated from all match records, or 10 matches before prediction. Then, the transformed datasets are split into two set, training sets and test set. The training set is from all Premier League matches throughout 2008/2009 season to 2018/2019 season plus the first 50 matches of 2019/2020 season. The test set will cover the remaining matches of 2019/2020 season before the league is suspended due to COVID-19 Outbreak. The accuracy of the best model is about 78%, which is pretty accurate for football match prediction. The results show that features that are made to quantify the performance of each team in the season or during the match is the features that are deemed to be important for the model. These features are recommended by the researcher to be used for further researches. |
|---|