|
|
--- |
|
|
license: mit |
|
|
language: |
|
|
- en |
|
|
pipeline_tag: tabular-classification |
|
|
library_name: sklearn |
|
|
tags: |
|
|
- traffic |
|
|
- sklearn |
|
|
- tabular-classification |
|
|
--- |
|
|
# TPnet-baseline |
|
|
|
|
|
TPnet-baseline is a Random Forest classifier trained on smart mobility and traffic features to predict traffic congestion levels (Low, Medium, High) in urban environments. |
|
|
|
|
|
## Model Details |
|
|
|
|
|
- **Model type**: Random Forest Classifier |
|
|
- **Input features**: 20 numerical features including vehicle count, road occupancy, weather, traffic light status, time-of-day, and more |
|
|
- **Output**: Multiclass classification – `High`, `Medium`, `Low` traffic congestion |
|
|
- **License**: MIT |
|
|
- **Trained on**: Smart Mobility Traffic Dataset from Kaggle |
|
|
|
|
|
## Training Details |
|
|
|
|
|
- Train/test split: 80/20 |
|
|
- Accuracy (test): 99.9% |
|
|
- F1 Score: 0.999 |
|
|
- Class-balanced via stratified sampling |
|
|
- No overfitting observed |
|
|
|
|
|
## Evaluation |
|
|
|
|
|
| Metric | Value | |
|
|
|------------|--------| |
|
|
| Accuracy | 99.9% | |
|
|
| F1 Score | 0.999 | |
|
|
| Model Size | ~1.2MB | |
|
|
|
|
|
Confusion matrix and full report are available in the repository. |
|
|
|
|
|
## How to Use |
|
|
|
|
|
```python |
|
|
import pickle |
|
|
|
|
|
with open("traffic_predictor_rf.pkl", "rb") as f: |
|
|
model = pickle.load(f) |
|
|
|
|
|
y_pred = model.predict(X_test) # where X_test is a [n_samples, 20] array |
|
|
``` |
|
|
|
|
|
## Limitations |
|
|
|
|
|
- Does not account for live data |
|
|
|
|
|
- Designed for offline batch inference |
|
|
|
|
|
- Assumes all 20 features are properly preprocessed and scaled |
|
|
|
|
|
|
|
|
## Authors |
|
|
|
|
|
- Created by [@Yukin3](https://github.com/Yukin3) |
