# 📊 Aduan Classification Model (IndoBERT)

Model ini dilatih untuk **klasifikasi teks aduan masyarakat** dalam Bahasa Indonesia menggunakan **IndoBERT (indobenchmark/indobert-base-p1)**.  
Model dapat mengelompokkan aduan ke dalam 4 kategori:

- **DARURAT** → Situasi darurat (kebakaran, kecelakaan, bencana)
- **PRIORITAS** → Perlu penanganan cepat (jalan rusak, kebersihan, infrastruktur)
- **UMUM** → Informasi / pertanyaan umum
- **LAINNYA** → Aduan lain yang tidak termasuk kategori di atas

---

## 📂 Files

- `model.safetensors` → model terlatih (498MB)
- `aduan_model.pt` → backup format pickle
- `config.json`, `tokenizer.json`, `vocab.txt` → konfigurasi dan tokenizer
- `special_tokens_map.json`, `tokenizer_config.json` → mapping tokenizer

---

## 📊 Dataset & Training

- **Total data (raw)**: 3,373  
  - Darurat: 900  
  - Prioritas: 875  
  - Umum: 880  
  - Lainnya: 718  
- **Augmentasi** → 3,600 (balance 900 per kelas)  
- **Split** → 80% Train (2880) | 20% Validation (720)  
- **Base model** → `indobenchmark/indobert-base-p1`  
- **Device training** → NVIDIA RTX 3050 Laptop GPU (CUDA)  

---

## 📈 Hasil Evaluasi

- **Best Epoch** → 3  
- **Validation Accuracy** → **93.89%**  
- **Macro F1-score** → **0.9389**

### 📑 Classification Report
| Label      | Precision | Recall | F1-score |
|------------|-----------|--------|----------|
| Darurat    | 0.9435    | 0.9278 | 0.9356   |
| Prioritas  | 0.9257    | 0.9000 | 0.9127   |
| Umum       | 0.9026    | 0.9778 | 0.9387   |
| Lainnya    | 0.9884    | 0.9500 | 0.9688   |
| **Macro Avg** | 0.9401 | 0.9389 | 0.9389   |

### 🔢 Confusion Matrix
```

[[167  10   3   0]   # Darurat
[  6 162  11   1]   # Prioritas
[  1   2 176   1]   # Umum
[  3   1   5 171]]  # Lainnya

````

---

## 🧪 Contoh Prediksi

### Single Input
```python
from transformers import AutoTokenizer, AutoModelForSequenceClassification
import torch

model_name = "Zulkifli1409/aduan-model"
tokenizer = AutoTokenizer.from_pretrained(model_name)
model = AutoModelForSequenceClassification.from_pretrained(model_name)

text = "Ada kebakaran besar di jalan sudirman, tolong kirim pemadam!"
inputs = tokenizer(text, return_tensors="pt", padding=True, truncation=True)
outputs = model(**inputs)
probs = torch.nn.functional.softmax(outputs.logits, dim=1)

pred_idx = torch.argmax(probs).item()
labels = ["DARURAT", "PRIORITAS", "UMUM", "LAINNYA"]

print("Prediksi:", labels[pred_idx])
print("Probabilitas:", probs.tolist())
````

### Output:

```
Prediksi: DARURAT
Probabilitas: [[0.9823, 0.0145, 0.0021, 0.0011]]
```

---

## 📦 Advanced Prediction Tests

| Teks Aduan                                | Prediksi  | Confidence |
| ----------------------------------------- | --------- | ---------- |
| ada kebakaran besar di pasar tolong cepat | DARURAT   | 60.62%     |
| jalan berlubang perlu diperbaiki          | PRIORITAS | 78.47%     |
| mohon pencerahan tentang program desa     | UMUM      | 72.09%     |
| ada orang kecelakaan parah butuh ambulans | DARURAT   | 74.29%     |
| sampah menumpuk di jalan                  | PRIORITAS | 71.17%     |
| banjir tinggi merendam rumah warga        | DARURAT   | 58.01%     |

---

## 🚀 Deployment

Model ini juga tersedia dalam bentuk API di Railway:

```
Base URL: https://api-klasifikasi-aduan.up.railway.app
```

Contoh request:

```bash
curl -X POST https://api-klasifikasi-aduan.up.railway.app/predict \
  -H "Content-Type: application/json" \
  -d '{"text": "Ada kebakaran di pasar"}'
```

Response:

```json
{
  "label": "DARURAT",
  "confidence": 0.9823,
  "all_scores": {
    "DARURAT": 0.9823,
    "PRIORITAS": 0.0145,
    "UMUM": 0.0021,
    "LAINNYA": 0.0011
  }
}
```

---

## 📧 Kontak

Dikembangkan oleh **Zulkifli1409**
Jika ada pertanyaan atau saran, silakan buka *issue* atau hubungi via [Hugging Face profile](https://huggingface.co/Zulkifli1409).

---

**© 2025 Klasifikasi Aduan Model**