factors

Unified factor library for quantitative finance, serving both alpha generation (tarifa) and risk modeling (perth) use cases.

Features

• Momentum Factors: Short-term (1mo), medium-term (6mo), and long-term (12mo) momentum
• Value Factors: Book-to-price, earnings yield, and free cash flow yield
• Quality Factors: ROE, ROA, profit margins, and leverage
• Size Factors: Log market capitalization
• Volatility Factors: Market beta and historical volatility
• Growth Factors: Earnings growth and sales growth
• Liquidity Factors: Turnover ratio and Amihud illiquidity

Architecture

factors/
├── traits.rs           # Core Factor trait definition
├── registry.rs         # Factor discovery and introspection
├── standardize.rs      # Cross-sectional z-scoring utilities
├── momentum/           # Trend persistence factors
├── value/              # Relative valuation factors
├── quality/            # Profitability and leverage factors
├── size/               # Market capitalization factors
├── volatility/         # Risk and beta factors
├── growth/             # Growth rate factors
└── liquidity/          # Trading volume factors

Usage

use factors::{Factor, FactorRegistry, FactorCategory};

// Create registry with all default factors
let registry = FactorRegistry::with_defaults();

// Get all momentum factors
let momentum = registry.by_category(FactorCategory::Momentum);

// Compute a specific factor
let short_momentum = registry.get("short_term_momentum").unwrap();
let result = short_momentum.compute(&data, date)?;

Factor Trait

All factors implement the core Factor trait:

pub trait Factor: Send + Sync {
    fn name(&self) -> &str;
    fn description(&self) -> &str;
    fn category(&self) -> FactorCategory;
    fn required_columns(&self) -> &[&str];
    fn lookback(&self) -> usize;
    fn compute_raw(&self, data: &LazyFrame, date: NaiveDate) -> Result<DataFrame>;
    fn compute(&self, data: &LazyFrame, date: NaiveDate) -> Result<DataFrame>;
}

Standardization

All factors support cross-sectional standardization:

• cross_sectional_standardize: Z-score normalization per date
• winsorize: Clip extreme values to percentiles
• robust_standardize: MAD-based standardization for outlier robustness

License

MIT License - see LICENSE.