|
|
--- |
|
|
tags: |
|
|
- pytorch |
|
|
- xG |
|
|
- football |
|
|
- soccer |
|
|
- expected-goals |
|
|
- mlp |
|
|
- binary-classification |
|
|
library_name: pytorch |
|
|
--- |
|
|
|
|
|
# MLP xG Prediction Model |
|
|
|
|
|
This is a Multi-Layer Perceptron (MLP) model trained to predict Expected Goals (xG) in football/soccer. |
|
|
|
|
|
## Model Description |
|
|
|
|
|
- **Architecture**: Multi-Layer Perceptron with 3 hidden layers |
|
|
- **Hidden Dimensions**: 128 → 64 → 32 |
|
|
- **Input Features**: 22 features |
|
|
- **Output**: Binary probability (goal vs no goal) |
|
|
- **Framework**: PyTorch |
|
|
- **Dropout Rate**: 0.3 |
|
|
- **Activation**: ReLU (hidden layers), Sigmoid (output) |
|
|
- **Normalization**: Batch Normalization after each hidden layer |
|
|
|
|
|
## Performance Metrics |
|
|
|
|
|
- **Accuracy**: 0.8797 |
|
|
- **Precision**: 0.6452 |
|
|
- **Recall**: 0.1225 |
|
|
- **F1 Score**: 0.2060 |
|
|
- **ROC AUC**: 0.7866 |
|
|
- **Log Loss**: 0.3169 |
|
|
|
|
|
## Features |
|
|
|
|
|
The model uses the following 22 features: |
|
|
|
|
|
- angle_to_gk |
|
|
- angle_to_goal |
|
|
- ball_closer_than_gk |
|
|
- body_part_name_Left Foot |
|
|
- body_part_name_Other |
|
|
- body_part_name_Right Foot |
|
|
- dist_to_gk |
|
|
- distance_to_goal |
|
|
- goal_dist_to_gk |
|
|
- minute |
|
|
- nearest_opponent_dist |
|
|
- nearest_teammate_dist |
|
|
- opponents_within_5m |
|
|
- play_pattern_name_From Counter |
|
|
- play_pattern_name_From Free Kick |
|
|
- play_pattern_name_From Goal Kick |
|
|
- play_pattern_name_From Keeper |
|
|
- play_pattern_name_From Kick Off |
|
|
- play_pattern_name_From Throw In |
|
|
- play_pattern_name_Other |
|
|
- play_pattern_name_Regular Play |
|
|
- teammates_within_5m |
|
|
|
|
|
## Usage |
|
|
|
|
|
```python |
|
|
import torch |
|
|
import joblib |
|
|
from huggingface_hub import hf_hub_download |
|
|
|
|
|
# Download files |
|
|
model_path = hf_hub_download(repo_id="rokati/mlp_xg", filename="best_mlp_model.pth") |
|
|
architecture_path = hf_hub_download(repo_id="rokati/mlp_xg", filename="model_architecture.py") |
|
|
scaler_path = hf_hub_download(repo_id="rokati/mlp_xg", filename="scaler.pkl") |
|
|
config_path = hf_hub_download(repo_id="rokati/mlp_xg", filename="config.json") |
|
|
|
|
|
# Load architecture |
|
|
import importlib.util |
|
|
spec = importlib.util.spec_from_file_location("model_architecture", architecture_path) |
|
|
model_module = importlib.util.module_from_spec(spec) |
|
|
spec.loader.exec_module(model_module) |
|
|
|
|
|
# Load model |
|
|
model = model_module.MLP(input_dim=22, hidden_dims=[128, 64, 32], dropout_rate=0.3) |
|
|
model.load_state_dict(torch.load(model_path, map_location=torch.device('cpu'))) |
|
|
model.eval() |
|
|
|
|
|
# Load scaler |
|
|
scaler = joblib.load(scaler_path) |
|
|
|
|
|
# Make prediction |
|
|
# X_new should be a pandas DataFrame or numpy array with the correct features |
|
|
X_scaled = scaler.transform(X_new) |
|
|
X_tensor = torch.FloatTensor(X_scaled) |
|
|
with torch.no_grad(): |
|
|
xg_prediction = model(X_tensor).numpy() |
|
|
``` |
|
|
|
|
|
## Training |
|
|
|
|
|
The model was trained on football shot event data with: |
|
|
- Binary Cross Entropy loss |
|
|
- Adam optimizer (lr=0.001, weight_decay=1e-5) |
|
|
- ReduceLROnPlateau scheduler |
|
|
- Batch size: 256 |
|
|
- Epochs: 50 |
|
|
|
|
|
## License |
|
|
|
|
|
MIT |
|
|
|
