---
license: apache-2.0
language:
- en
metrics:
- accuracy
pipeline_tag: tabular-classification
library_name: sklearn
tags:
- stacking-ensemble
- xgboost
- catboost
- lightgbm
- adaboost
- randomforest
model-index:
- name: stacking-ensemble-learning
  results:
  - task:
      type: tabular-classification
      name: Network Intrusion Detection
    dataset:
      name: CSE-CIC-IDS2018 Cleaned
      type: cicids2018
      url: https://www.kaggle.com/datasets/ekkykharismadhany/csecicids2018-cleaned
    metrics:
    - type: accuracy
      value: 1.0
      name: Accuracy (Meta Model)
    - type: f1
      value: 1.0
      name: F1 Score (Meta Model)
    - type: precision
      value: 1.0
      name: Precision (Meta Model)
    - type: recall
      value: 1.0
      name: Recall (Meta Model)
---

# IDS Stacking Ensemble Learning

Stacking ensemble (XGBoost + CatBoost + LightGBM + AdaBoost) for base mdoel with Random Forest as meta-model for Network Intrusion Detection System (IDS).

## Metrics (test set)

| Model | Accuracy | F1 |
|-------|----------|----|
| XGBoost | 1.0000 | 1.0000 |
| CatBoost | 0.9998 | 0.9998 |
| LightGBM | 0.4441 | 0.5403 |
| AdaBoost | 0.9911 | 0.9907 |

## Requirements

```bash
pip install scikit-learn xgboost catboost lightgbm pandas numpy joblib huggingface_hub
```

## Usage

```python
import joblib
import numpy as np
from huggingface_hub import hf_hub_download

models = {k: joblib.load(hf_hub_download("mrsindhunugroho/stacking-ensemble-learning", f"models/{k}_model.pkl"))
          for k in ["xgboost", "catboost", "lightgbm", "adaboost"]}
meta = joblib.load(hf_hub_download("mrsindhunugroho/stacking-ensemble-learning", "models/meta_model.pkl"))
le   = joblib.load(hf_hub_download("mrsindhunugroho/stacking-ensemble-learning", "models/label_encoder.pkl"))

base_preds = np.column_stack([m.predict(X) for m in models.values()])
y_pred = le.inverse_transform(meta.predict(base_preds))
```

## Dataset

- **Source:** Kaggle — [CSE-CIC-IDS2018 Cleaned](https://www.kaggle.com/datasets/ekkykharismadhany/csecicids2018-cleaned)
- **Original:** Canadian Institute for Cybersecurity (CSE-CIC-IDS2018)
- **Preprocessing:** sampling, label encoding, imputation, feature sanitization

## Author

Sindhu Nugroho — [ORCID](https://orcid.org/0009-0002-1558-4574)