"Transform Trading Data into Actionable Intelligence"
Aplikasi analisis probabilitas trading berbasis Dash Plotly yang mengubah data historis trading menjadi insight probabilistik untuk pengambilan keputusan yang lebih baik.
**Demo APP : https://analis.bamsbung.id/
Trading bukan tentang memprediksi masa depan dengan pasti, tetapi tentang memahami probabilitas dan mengelola risiko. Aplikasi ini dibangun dengan filosofi:
-
** Data-Driven Decision Making**
- Setiap keputusan harus didukung oleh data historis
- Probabilitas kondisional lebih akurat dari prediksi absolut
- Pattern recognition melalui analisis statistik
-
** Embrace Uncertainty**
- Tidak ada strategi yang menang 100%
- Yang penting adalah positive expectancy dalam jangka panjang
- Risk management lebih penting dari win rate
-
** Scientific Approach**
- Hypothesis testing dengan data
- Validasi model dengan calibration
- Continuous improvement melalui feedback loop
-
** Risk-Reward Balance**
- Fokus pada R-multiple, bukan profit absolut
- Optimize expectancy, bukan win rate
- Position sizing berdasarkan Kelly Criterion
Trading penuh dengan noise dan emosi. Aplikasi ini memberikan clarity melalui:
- Visualisasi probabilitas yang mudah dipahami
- Identifikasi kondisi pasar terbaik untuk trading
- Deteksi pattern yang tidak terlihat dengan mata telanjang
Bukan hanya angka dan grafik, tetapi rekomendasi konkret:
- "Trade only when composite score > 70"
- "Optimal SL at 0.8R based on MAE analysis"
- "Use 0.5% risk per trade (Half Kelly) for this strategy"
Memahami worst-case scenario sebelum terjadi:
- Monte Carlo simulation menunjukkan kemungkinan drawdown maksimal
- Risk of ruin calculator mencegah over-leverage
- Expectancy analysis mengidentifikasi strategi negatif
Framework untuk iterasi dan optimasi:
- Backtest berbagai threshold dan parameter
- Compare multiple scenarios side-by-side
- Track improvement over time
"Transform Trading Data into Actionable Intelligence"
Aplikasi analisis probabilitas trading berbasis Dash Plotly dengan 18 halaman analisis komprehensif untuk mengubah data historis trading menjadi insight probabilistik.
| Nilai | Manfaat Konkret |
|---|---|
| Clarity | Visualisasi probabilitas yang mudah dipahami |
| Actionable | Rekomendasi konkret untuk trading |
| Risk Awareness | Monte Carlo simulation, risk of ruin analysis |
| Efficiency | Dari 3 hari analisis manual → 5 menit otomatis |
| # | Halaman | Fungsi Utama |
|---|---|---|
| 1 | Trade Analysis Dashboard | Overview performa, expectancy, equity curve |
| 2 | Probability Explorer | Probabilitas kondisional, composite score |
| 3 | Sequential Analysis | Markov chain, streak analysis |
| 4 | Calibration Lab | Reliability diagram, Brier score |
| 5 | Regime Explorer | Performa per kondisi pasar |
| 6 | What-If Scenarios | Monte Carlo, MAE/MFE optimizer |
| 7 | Auto Feature Selection ⭐ | ML-based feature ranking (30 fitur → 8 optimal) |
| 8 | Market Condition Scoring ⭐ | Dual mode: v1 Trade Regime + v2 Market State |
| 9 | ML Prediction Engine ⭐ | Calibrated probability + Conformal Prediction + Decision Engine |
| 10 | Combination Probability Analyzer ⭐ | Bayesian analysis + SQA Optimizer + Cross-validation |
| 11 | Decision Tree Rules ⭐ | Extract IF-THEN rules untuk MQL5 EA |
| 12 | Walk-Forward Validation ⭐ | Time-series cross-validation |
| 13 | Ensemble Voting ⭐ | Multi-method consensus voting |
| 14 | Optuna Hyperparameter Optimizer ⭐ | Bayesian optimization untuk ML pipeline |
| 15 | Isolation Forest Analyzer | Deteksi positive outliers dalam winning trades |
| 16 | NGBoost Probabilistic Engine | Native uncertainty estimation dengan distribusi probabilitas |
| 17 | Risk Lab | Risk OS, policy builder, position sizing, gating rules |
| 18 | Neural Network Lab 🆕 | MLP, LSTM, Transformer + ONNX Export untuk MT5 |
| Feature | Deskripsi |
|---|---|
| Server DataFrame Migration | Final validation complete, 253,327x speedup |
| Market Scoring Bug Fixes | Fixed n_samples=0 dan KeyError: 'score' |
| Send to Walk-Forward 🆕 | Tombol baru di Page 9 untuk kirim prediksi ke Page 12 |
| Performance Benchmark | Script benchmark untuk migration metrics |
| Feature | Deskripsi |
|---|---|
| Neural Network Lab (Page 18) 🆕 | Deep learning dengan MLP, LSTM, Transformer |
| ONNX Export 🆕 | Export neural network models ke ONNX untuk MT5 EA |
| Ensemble NN 🆕 | Kombinasi multiple NN models dengan weighted averaging |
| Uncertainty Estimation 🆕 | MC Dropout untuk confidence levels |
| Batch Prediction 🆕 | Process semua trades sekaligus dengan histogram |
| Risk Lab (Page 17) | Risk OS dengan policy builder, hard gating, soft throttling |
| NGBoost Engine (Page 16) | Native probabilistic predictions dengan distribusi uncertainty |
# 1. Install dependencies
pip install -r requirements.txt
# 2. Run application
python app.py
# 3. Open browser
http://127.0.0.1:8050Load Data → Auto Feature Selection (Page 7) → Market Condition Scoring (Page 8)
↓
Optuna Optimizer (Page 14) → Optimize hyperparameters & features
↓
Isolation Forest (Page 15) → Deteksi positive outliers → Enrich ke Page 9
↓
Combination Analyzer + SQA (Page 10) → Decision Tree Rules (Page 11)
↓
Walk-Forward Validation (Page 12) → Ensemble Voting (Page 13)
↓
ML Prediction Engine (Page 9) ←→ NGBoost Engine (Page 16)
↓
Neural Network Lab (Page 18) → Train MLP/LSTM/Transformer → ONNX Export 🆕
↓
Risk Lab (Page 17) → Risk policy, sizing, gating → EA Export ✅
| Dokumen | Deskripsi |
|---|---|
| README_FULL.md | Dokumentasi lengkap & komprehensif |
| ARCHITECTURE.md | Arsitektur sistem & data flow |
| PANDUAN_LENGKAP.md | Panduan penggunaan setiap halaman |
| Version.md | Version history & roadmap |
- Frontend: Dash, Plotly, Bootstrap
- Backend: Python, Pandas, NumPy
- ML: LightGBM, NGBoost, Scikit-learn, SHAP
- Deep Learning: TensorFlow/Keras, ONNX 🆕
- Optimization: Optuna (Bayesian), SQA (Quantum Annealing)
- Tracking: MLflow
- Testing: Pytest, Hypothesis (Property-Based Testing)
Built with ❤️ for traders who believe in data-driven decisions.
Version 5.1 | 26 December 2025
ML Prediction Engine Status: ✅ PRODUCTION READY
Ready to predict. Ready to win.