# nt-features
[](https://crates.io/crates/nt-features)
[](https://docs.rs/nt-features)
[](LICENSE)
[](https://www.rust-lang.org/)
**Technical indicators and feature engineering for quantitative trading strategies.**
The `nt-features` crate provides a comprehensive library of technical indicators, statistical features, and feature engineering tools optimized for algorithmic trading.
## Features
- **100+ Technical Indicators** - Moving averages, oscillators, volatility, volume
- **Statistical Features** - Rolling statistics, correlations, distributions
- **Feature Engineering** - Automated feature generation and selection
- **Polars Integration** - High-performance DataFrame operations
- **Lazy Evaluation** - Efficient computation with Polars lazy API
- **Custom Indicators** - Extensible framework for custom indicators
- **Real-Time Processing** - Stream-friendly incremental calculations
- **GPU Acceleration** - Optional CUDA support for large datasets
## Technical Indicators
### Trend Indicators
- Simple Moving Average (SMA)
- Exponential Moving Average (EMA)
- Weighted Moving Average (WMA)
- MACD (Moving Average Convergence Divergence)
- ADX (Average Directional Index)
- Parabolic SAR
### Momentum Indicators
- RSI (Relative Strength Index)
- Stochastic Oscillator
- Williams %R
- Rate of Change (ROC)
- Momentum
- CCI (Commodity Channel Index)
### Volatility Indicators
- Bollinger Bands
- ATR (Average True Range)
- Keltner Channels
- Standard Deviation
- Historical Volatility
### Volume Indicators
- On-Balance Volume (OBV)
- Volume-Weighted Average Price (VWAP)
- Accumulation/Distribution
- Chaikin Money Flow
- Volume Oscillator
## Quick Start
```toml
[dependencies]
nt-features = "0.1"
```
### Basic Usage
```rust
use nt_features::{
indicators::{SMA, RSI, BollingerBands, MACD},
FeatureEngine,
};
use polars::prelude::*;
#[tokio::main]
async fn main() -> anyhow::Result<()> {
// Load market data
let df = CsvReader::from_path("AAPL.csv")?
.has_header(true)
.finish()?;
// Calculate indicators
let sma_20 = SMA::new(20).calculate(&df)?;
let rsi_14 = RSI::new(14).calculate(&df)?;
let bb = BollingerBands::new(20, 2.0).calculate(&df)?;
let macd = MACD::new(12, 26, 9).calculate(&df)?;
// Add indicators to DataFrame
let df = df
.lazy()
.with_column(sma_20.alias("sma_20"))
.with_column(rsi_14.alias("rsi_14"))
.with_column(bb.upper.alias("bb_upper"))
.with_column(bb.lower.alias("bb_lower"))
.with_column(macd.macd_line.alias("macd"))
.with_column(macd.signal_line.alias("macd_signal"))
.collect()?;
println!("{}", df);
Ok(())
}
```
### Feature Engineering
```rust
use nt_features::{FeatureEngine, FeatureConfig};
use polars::prelude::*;
#[tokio::main]
async fn main() -> anyhow::Result<()> {
let config = FeatureConfig {
lookback_periods: vec![5, 10, 20, 50, 200],
include_technical: true,
include_statistical: true,
include_interactions: true,
pca_components: Some(50),
};
let engine = FeatureEngine::new(config);
// Load and prepare data
let df = CsvReader::from_path("market_data.csv")?
.has_header(true)
.finish()?;
// Generate features automatically
let features = engine.generate_features(&df)?;
println!("Generated {} features", features.width());
println!("Feature names: {:?}", features.get_column_names());
// Feature importance
let importance = engine.calculate_importance(&features, "returns")?;
println!("Top 10 features:\n{}", importance.head(Some(10)));
Ok(())
}
```
### Real-Time Indicators
```rust
use nt_features::indicators::streaming::{StreamingSMA, StreamingRSI};
#[tokio::main]
async fn main() -> anyhow::Result<()> {
let mut sma = StreamingSMA::new(20);
let mut rsi = StreamingRSI::new(14);
// Process streaming data
let prices = vec![100.0, 101.5, 99.8, 102.3, 103.1];
for price in prices {
let sma_value = sma.update(price);
let rsi_value = rsi.update(price);
println!("Price: {}, SMA: {:.2}, RSI: {:.2}", price, sma_value, rsi_value);
}
Ok(())
}
```
## Architecture
```
nt-features/
├── indicators/
│ ├── trend.rs # Trend indicators (SMA, EMA, MACD)
│ ├── momentum.rs # Momentum indicators (RSI, Stochastic)
│ ├── volatility.rs # Volatility indicators (Bollinger, ATR)
│ ├── volume.rs # Volume indicators (OBV, VWAP)
│ └── streaming.rs # Real-time streaming indicators
├── statistics/
│ ├── rolling.rs # Rolling statistics
│ ├── correlation.rs # Correlation analysis
│ └── distributions.rs # Distribution metrics
├── feature_engine.rs # Automated feature generation
├── selection.rs # Feature selection algorithms
└── lib.rs
```
## Performance
| SMA calculation | 10M+ samples/sec | <1μs |
| RSI calculation | 5M+ samples/sec | <2μs |
| Feature generation | 100K+ rows/sec | <10ms |
| Streaming update | <100ns | Real-time |
Benchmarks run on: AMD EPYC 7763 (single core)
## Dependencies
| `nt-core` | Core types |
| `polars` | DataFrame operations |
| `statrs` | Statistical functions |
| `ndarray` | Numerical arrays |
| `rayon` | Parallel processing |
## Testing
```bash
# Unit tests
cargo test -p nt-features
# Benchmarks
cargo bench -p nt-features
# Property-based tests
cargo test -p nt-features --features proptest
```
## Contributing
Contributions welcome! Areas for improvement:
- Additional technical indicators
- GPU acceleration for more indicators
- Feature selection algorithms
- Documentation and examples
## License
Licensed under MIT OR Apache-2.0.