---
language:
- id
tags:
- text-classification
- emotion-detection
- bilstm
- bahasa-indonesia
- tenang-in
license: mit
---

# Model 1 — Klasifikasi Emosi (Tenang.in)

Model BiLSTM untuk klasifikasi 7 emosi dari teks jurnal harian bahasa Indonesia.
Bagian dari sistem deteksi dini burnout **Tenang.in**.

## Performa

| Metrik | Nilai |
|--------|-------|
| Test Accuracy | 93.75% |
| Test F1 Weighted | 0.9377 |
| Jumlah kelas | 7 emosi |

## Label

| ID | Label |
|----|-------|
| 0 | anger |
| 1 | anticipation |
| 2 | disgust |
| 3 | fear |
| 4 | joy |
| 5 | sadness |
| 6 | trust |

## Artefak

| File | Kegunaan |
|------|----------|
| model1_bilstm_final.keras | Model TensorFlow |
| tokenizer_final.pkl | Keras Tokenizer |
| w2v_final.bin | Word2Vec embeddings |
| config.json | Konfigurasi model |

## Arsitektur
Input (50 token) → Embedding (Word2Vec pretrained)
→ BiLSTM(64) → BiLSTM(32) → GlobalMaxPool
→ Dense(64) → Output softmax (7 kelas)
Total params: 422.211 (1.61 MB)

## Penggunaan

```python
from inference import EmotionPredictor

predictor = EmotionPredictor()
result    = predictor.predict("aku sangat lelah kerja terus menerus")
print(result['label'])       # sadness
print(result['confidence'])  # 0.918
```