GifariKemalandClaude Opus 4.6
e8355b3f62
- Dark mode: class-based theme toggle with localStorage persistence and flash prevention - Trade History (/trades): paginated table, stats cards, equity curve chart with DB API endpoints - Backtest Viewer (/backtests): log parser for 35 backtest results, sidebar + detail + comparison tabs - Model Insights: dashboard card + dialog showing feature importance, regime distribution, training history - Alert/Signal Log (/alerts): signal stats, filterable table with execution tracking - API: 8 new endpoints with psycopg2 DB connection pool - Dark mode sweep across books page, about dialog, and all dashboard components - Architecture docs rewritten with Mermaid diagrams (23 docs) - README and FEATURES.md rewritten bilingual (Indonesian + English) - main_live.py: write model_metrics.json on startup and retrain Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
136 lines
3.8 KiB
Markdown
136 lines
3.8 KiB
Markdown
# HMM (*Hidden Markov Model*) — *Regime Detector*
|
|
|
|
> **File:** `src/regime_detector.py`
|
|
> **Model:** `models/hmm_regime.pkl`
|
|
> **Library:** `hmmlearn.GaussianHMM`
|
|
|
|
---
|
|
|
|
## Apa Itu HMM?
|
|
|
|
*Hidden Markov Model* (HMM) adalah model statistik yang mengidentifikasi **kondisi tersembunyi** (*hidden states*) dari data yang dapat diamati. Dalam konteks *trading*, HMM mendeteksi **3 kondisi pasar** (*regime*) yang tidak terlihat langsung dari harga:
|
|
|
|
```mermaid
|
|
graph LR
|
|
A["Data Pasar<br/>Return, Volatilitas, Volume"] --> B["HMM<br/>GaussianHMM 3-state"]
|
|
B --> C["Low Volatility<br/>🟢 TRADE"]
|
|
B --> D["Medium Volatility<br/>🟡 REDUCE"]
|
|
B --> E["High Volatility<br/>🔴 SLEEP"]
|
|
```
|
|
|
|
---
|
|
|
|
## 3 *State* Pasar
|
|
|
|
| *State* | Label | Rekomendasi | Efek pada *Trading* |
|
|
|---------|-------|-------------|---------------------|
|
|
| **0** | *Low Volatility* | **TRADE** | *Lot* normal, semua filter aktif |
|
|
| **1** | *Medium Volatility* | **REDUCE** | *Lot* dikurangi, *entry* lebih ketat |
|
|
| **2** | *High Volatility* / Krisis | **SLEEP** | **Tidak boleh *trading*** — terlalu berisiko |
|
|
|
|
---
|
|
|
|
## Cara Kerja
|
|
|
|
### *Input Features* (3 fitur)
|
|
|
|
```python
|
|
features = [
|
|
"returns", # Perubahan harga (%)
|
|
"volatility", # Volatilitas rolling (standar deviasi)
|
|
"volume_change" # Perubahan volume (%)
|
|
]
|
|
```
|
|
|
|
### Proses *Training*
|
|
|
|
```python
|
|
class MarketRegimeDetector:
|
|
def __init__(self,
|
|
n_regimes=3, # 3 state
|
|
lookback_periods=500, # 500 bar untuk training
|
|
retrain_frequency=20, # Retrain setiap 20 bar baru
|
|
covariance_type="full",
|
|
random_state=42,
|
|
):
|
|
self.hmm = GaussianHMM(
|
|
n_components=3,
|
|
covariance_type="full",
|
|
n_iter=100,
|
|
random_state=42,
|
|
)
|
|
```
|
|
|
|
### Proses Deteksi
|
|
|
|
```python
|
|
# 1. Siapkan data 500 bar terakhir
|
|
X = df[["returns", "volatility", "volume_change"]].to_numpy()
|
|
|
|
# 2. Fit model (atau load dari .pkl)
|
|
self.hmm.fit(X)
|
|
|
|
# 3. Prediksi state saat ini
|
|
state = self.hmm.predict(X)[-1] # State terakhir
|
|
|
|
# 4. Hitung probabilitas tiap state
|
|
probs = self.hmm.predict_proba(X)[-1]
|
|
# Contoh: [0.85, 0.10, 0.05] = 85% low vol
|
|
```
|
|
|
|
---
|
|
|
|
## Output: `RegimeState`
|
|
|
|
```python
|
|
@dataclass
|
|
class RegimeState:
|
|
regime: MarketRegime # LOW/MEDIUM/HIGH_VOLATILITY atau CRISIS
|
|
confidence: float # Probabilitas state terpilih (0-1)
|
|
probabilities: Dict # Probabilitas semua state
|
|
volatility: float # Level volatilitas saat ini
|
|
recommendation: str # "TRADE", "REDUCE", atau "SLEEP"
|
|
```
|
|
|
|
---
|
|
|
|
## Integrasi dengan Sistem
|
|
|
|
```mermaid
|
|
graph TD
|
|
A["HMM Regime Detector"] --> B{"Regime?"}
|
|
B -->|LOW VOL| C["✅ TRADE<br/>Lot normal, semua filter aktif"]
|
|
B -->|MEDIUM VOL| D["⚠️ REDUCE<br/>Lot dikurangi, entry lebih ketat"]
|
|
B -->|HIGH VOL| E["🛑 SLEEP<br/>Blokir semua entry baru"]
|
|
C --> F["Entry Filter #2"]
|
|
D --> F
|
|
E -->|Blokir| G["Skip — tidak boleh trading"]
|
|
```
|
|
|
|
**Penggunaan dalam *main_live.py*:**
|
|
- *Regime* **SLEEP** → blokir semua *entry* baru (Filter #2)
|
|
- *Regime* memengaruhi *lot sizing* — `SmartRiskManager` mengurangi *lot* pada *medium volatility*
|
|
- *Regime* dicatat di setiap *trade log* untuk analisis historis
|
|
|
|
---
|
|
|
|
## Penyimpanan Model
|
|
|
|
- **Format:** `.pkl` (*pickle*)
|
|
- **Lokasi:** `models/hmm_regime.pkl`
|
|
- **Ukuran:** ~50-100 KB
|
|
- ***Retrain*:** Otomatis setiap 20 bar baru ATAU melalui `AutoTrainer` setiap 7 hari
|
|
- ***Auto-retrain* dipicu juga saat:** Akurasi deteksi turun atau distribusi *return* berubah signifikan
|
|
|
|
---
|
|
|
|
## Konfigurasi
|
|
|
|
Dari `src/config.py` → `RegimeConfig`:
|
|
|
|
| Parameter | Nilai | Keterangan |
|
|
|-----------|-------|------------|
|
|
| `n_regimes` | **3** | Jumlah *state* HMM |
|
|
| `lookback_periods` | **500** | Bar untuk *training* HMM |
|
|
| `retrain_frequency` | **20** | *Retrain* setiap 20 bar baru |
|