|
|
--- |
|
|
license: mit |
|
|
language: |
|
|
- en |
|
|
metrics: |
|
|
- accuracy |
|
|
- precision |
|
|
- recall |
|
|
- f1 |
|
|
- roc_auc |
|
|
pipeline_tag: tabular-classification |
|
|
tags: |
|
|
- classification |
|
|
- traffic |
|
|
--- |
|
|
|
|
|
# Model Card for Infinitode/TAPM-OPEN-ARC |
|
|
|
|
|
Repository: https://github.com/Infinitode/OPEN-ARC/ |
|
|
|
|
|
## Model Description |
|
|
|
|
|
OPEN-ARC-TAP is a straightforward XGBClassifier model developed as part of Infinitode's OPEN-ARC initiative. It was developed to assess the probability of traffic accidents by analyzing various external factors. |
|
|
|
|
|
**Architecture**: |
|
|
|
|
|
- **XGBClassifier**: `random_state=42`, `use_label_encoder=False`, `eval_metric='logloss'`, `colsample_bytree=0.8`, `learning_rate=0.01`, `max_depth=5`, `n_estimators=100`, `scale_pos_weight=1`, `subsample=0.8`. |
|
|
- **Framework**: XGBoost |
|
|
- **Training Setup**: Trained without extra training params. |
|
|
|
|
|
## Uses |
|
|
|
|
|
- Identifying potential accident-prone or high-risk areas. |
|
|
- Enhancing preventive measures for traffic accidents and improving road safety. |
|
|
- Researching traffic safety. |
|
|
|
|
|
## Limitations |
|
|
|
|
|
- May produce implausible or inappropriate results when affected by extreme outlier values. |
|
|
- Might offer inaccurate predictions regarding the likelihood of an accident; caution is recommended when interpreting these outputs. |
|
|
|
|
|
## Training Data |
|
|
|
|
|
- Dataset: Traffic Accident Prediction 💥🚗 dataset from Kaggle. |
|
|
- Source URL: https://www.kaggle.com/datasets/denkuznetz/traffic-accident-prediction |
|
|
- Content: Weather conditions, road types, time of day, and other factors, along with the occurrence or absence of an accident. |
|
|
- Size: 798 entries of traffic data. |
|
|
- Preprocessing: Mapped all string values to numeric values and dropped missing values. SMOTE was used to balance class imbalances. |
|
|
|
|
|
## Training Procedure |
|
|
|
|
|
- Metrics: accuracy, precision, recall, F1, ROC-AUC |
|
|
- Train/Testing Split: 80% train, 20% testing. |
|
|
|
|
|
## Evaluation Results |
|
|
|
|
|
| Metric | Value | |
|
|
| ------ | ----- | |
|
|
| Testing Accuracy | 85.2% | |
|
|
| Testing Weighted Average Precision | 87% | |
|
|
| Testing Weighted Average Recall | 85% | |
|
|
| Testing Weighted Average F1 | 85% | |
|
|
| Testing ROC-AUC | 82.5% | |
|
|
|
|
|
## How to Use |
|
|
|
|
|
```python |
|
|
import random |
|
|
|
|
|
def test_random_samples(model, X_test, y_test, n_samples=5): |
|
|
""" |
|
|
Selects random samples from the test set, makes predictions, and compares with actual values. |
|
|
|
|
|
Parameters: |
|
|
- model: Trained XGBoost classifier. |
|
|
- X_test: Feature set for testing. |
|
|
- y_test: True labels for testing. |
|
|
- n_samples: Number of random samples to test. |
|
|
|
|
|
Returns: |
|
|
None |
|
|
""" |
|
|
# Convert X_test and y_test to DataFrame for easier indexing |
|
|
X_test_df = X_test.reset_index(drop=True) |
|
|
y_test_df = y_test.reset_index(drop=True) |
|
|
|
|
|
# Pick random indices |
|
|
random_indices = random.sample(range(len(X_test)), n_samples) |
|
|
|
|
|
print("Testing on Random Samples:") |
|
|
for idx in random_indices: |
|
|
sample = X_test_df.iloc[idx] |
|
|
true_label = y_test_df.iloc[idx] |
|
|
|
|
|
# Predict using the model |
|
|
prediction = model.predict(sample.values.reshape(1, -1)) |
|
|
|
|
|
# Output results |
|
|
print(f"Sample Index: {idx}") |
|
|
print(f"Features: {sample.values}") |
|
|
print(f"True Label: {true_label}, Predicted Label: {prediction[0]}") |
|
|
print("-" * 40) |
|
|
|
|
|
# Example usage |
|
|
test_random_samples(xgb, X_test, y_test) |
|
|
``` |
|
|
|
|
|
## Contact |
|
|
|
|
|
For questions or issues, open a GitHub issue or reach out at https://infinitode.netlify.app/forms/contact. |
