Hybrid LSTM + Random Forest ensemble achieving 85%+ directional prediction accuracy — a 17% lift over the moving-average baseline across 10+ financial instruments.
Retail investors need directional signals from market data without writing code. Existing tools either require quantitative expertise or give generic, untimely signals. FinVizaard closes that gap with a production-grade ML pipeline and an interactive dashboard.
| Metric | Score |
|---|---|
| Directional Accuracy | 85%+ |
| Improvement over baseline | +17% |
| Instruments covered | 10+ |
| Technical indicators engineered | 12+ |
| Historical data span | 5+ years |
| Evaluation metrics | MAE, RMSE, Directional Accuracy |
Raw Market Data (5+ years) ↓ Data Cleaning & Normalization ↓ Feature Engineering (12+ indicators) RSI · MACD · Bollinger Bands · Moving Averages · Volume · Volatility ↓ LSTM + Random Forest Ensemble ↳ GridSearchCV Hyperparameter Tuning ↳ K-Fold Cross-Validation ↓ Model Evaluation (MAE · RMSE · Directional Accuracy) ↓ Interactive Dashboard (10+ visualizations) Trends · Volatility Bands · Confidence Intervals
| Layer | Technology |
|---|---|
| ML Models | LSTM, Random Forest, Ensemble Methods |
| Framework | TensorFlow, Scikit-learn |
| Data Processing | Pandas, NumPy, SciPy |
| Visualization | Matplotlib, Seaborn |
| Backend API | FastAPI |
| Frontend | React.js / Next.js |
| Database | MySQL |
git clone https://github.com/prince-pokharna/FinVizaard.git
cd FinVizaard
pip install -r requirements.txt
cd Backend/src
python main.py| Indicator | Type | Purpose |
|---|---|---|
| RSI (14-period) | Momentum | Overbought/oversold detection |
| MACD | Trend | Momentum crossover signals |
| Bollinger Bands | Volatility | Price deviation from mean |
| SMA (20, 50, 200) | Trend | Moving average baselines |
| EMA (12, 26) | Trend | Exponential smoothing |
| Volume Moving Avg | Volume | Liquidity signals |
| ATR | Volatility | True range measure |
Prince Pokharna
- GitHub: @prince-pokharna
- Email: prince187p0kharna@gmail.com
- LinkedIn: prince-pokharna