This was my attempt at building a model to predict English Premier League match outcomes (Home Win, Draw, Away Win). This was a much more complicated task than I first thought. I quickly realized how much feature engineering actually matters and how much information you can unlock if done correctly. I also got to see how effective AutoML can be for modelling data as well as, optimizing models without manually tweaking everything.
- Loaded multiple seasons of EPL data.
- Made sure everything was consistent across datasets.
- Built new features based on team form, head-to-head results, goal differences, and other match stats.
- Tested different machine learning models.
- Used AutoML to find the best-performing model.
- Measured performance with accuracy and other metrics.
- Looked at which features actually mattered in making predictions.
- Improve feature selection.
- Try other modeling approaches.
- Maybe bring in real-time betting odds to see if that improves results.