Struct HMMRegimeDetector 

pub struct HMMRegimeDetector { /* private fields */ }

Hidden Markov Model for regime detection.

Uses a 3-state HMM (by default) to model market regimes:

• State 0: Bull market (positive returns, low volatility)
• State 1: Bear market (negative returns, medium volatility)
• State 2: High volatility (any direction, high volatility)

The model uses the forward algorithm for online filtering and periodically re-estimates parameters using the Baum-Welch algorithm.

Example

use indicators::{HMMRegimeDetector, HMMConfig, MarketRegime};

let mut detector = HMMRegimeDetector::crypto_optimized();

// Feed close prices
for i in 0..200 {
    let price = 100.0 * (1.0 + 0.001 * i as f64); // gentle uptrend
    let result = detector.update(price);
    if detector.is_ready() {
        println!("HMM regime: {} (conf: {:.0}%)", result.regime, result.confidence * 100.0);
    }
}

Implementations

impl HMMRegimeDetector

pub fn new(config: HMMConfig) -> Self

Create a new HMM detector with the given configuration

pub fn default_config() -> Self

Create with default config

pub fn crypto_optimized() -> Self

Create optimized for crypto

pub fn conservative() -> Self

Create with conservative config

pub fn update(&mut self, close: f64) -> RegimeConfidence

Update with new close price and get regime.

Calculates log return from the previous close, then runs the forward algorithm step and optional parameter updates.

pub fn update_ohlc( &mut self, _high: f64, _low: f64, close: f64, ) -> RegimeConfidence

Update with OHLC data (uses close price for HMM)

pub fn get_regime_confidence(&self) -> RegimeConfidence

Get current regime with confidence

pub fn state_probabilities(&self) -> &[f64]

Get state probabilities

pub fn state_parameters(&self) -> Vec<(f64, f64)>

Get state parameters (mean, variance) for inspection

pub fn transition_matrix(&self) -> &[Vec<f64>]

Get transition matrix

pub fn current_state_index(&self) -> usize

Get current state index

pub fn is_ready(&self) -> bool

Check if model is warmed up (has enough observations)

pub fn expected_regime_duration(&self, state: usize) -> f64

Get expected regime duration (from transition matrix).

Expected duration = 1 / (1 - P(stay in state))

pub fn predict_next_state(&self) -> (usize, f64)

Predict most likely next state

pub fn n_observations(&self) -> usize

Get the total number of observations processed

pub fn current_confidence(&self) -> f64

Get the current confidence score

pub fn config(&self) -> &HMMConfig

Get the configuration

Trait Implementations

impl Debug for HMMRegimeDetector

fn fmt(&self, f: &mut Formatter<'_>) -> Result

Formats the value using the given formatter.