Enum HybridClassifier 

pub enum HybridClassifier {
    RuleBased(RuleBasedClassifier),
    Hybrid {
        ml_model: Box<TrainedModel>,
        fallback: RuleBasedClassifier,
        confidence_threshold: f32,
    },
}

Hybrid classifier combining rule-based and ML approaches

NLP-010: Integrates trained ML models with fallback to rule-based classification Implements three-tier architecture from nlp-models-techniques-spec.md:

• Tier 1: Rule-based (fast, <10ms)
• Tier 2: TF-IDF + Random Forest (medium, <100ms)
• Tier 3: Transformer models (future work)

Variants

RuleBased(RuleBasedClassifier)

Rule-based only (Tier 1)

Hybrid

ML model with rule-based fallback (Tier 2 + Tier 1)

Fields

ml_model: Box<TrainedModel>

fallback: RuleBasedClassifier

confidence_threshold: f32

Implementations

impl HybridClassifier

pub fn new_rule_based() -> Self

Create a new rule-based classifier (Tier 1 only)

Examples

use organizational_intelligence_plugin::classifier::HybridClassifier;

let classifier = HybridClassifier::new_rule_based();

pub fn new_hybrid(ml_model: TrainedModel, confidence_threshold: f32) -> Self

Load a trained ML model with rule-based fallback (Tier 2 + Tier 1)

Arguments

• model - Trained ML model
• confidence_threshold - Minimum confidence to use ML prediction (default: 0.60)

Examples

let classifier = HybridClassifier::new_hybrid(model, 0.65);

pub fn classify_from_message(&self, message: &str) -> Option<Classification>

Classify a commit message

Uses ML model if available and confident, otherwise falls back to rule-based.

Arguments

• message - Commit message to classify

Returns

• Some(Classification) if a category is detected
• None if no patterns match

Examples

use organizational_intelligence_plugin::classifier::HybridClassifier;

let classifier = HybridClassifier::new_rule_based();
if let Some(result) = classifier.classify_from_message("fix: null pointer") {
    println!("Category: {:?}, Confidence: {:.2}", result.category, result.confidence);
}

pub fn classify_multi_label( &self, message: &str, top_n: usize, min_confidence: f32, ) -> Result<MultiLabelClassification>

Classify with multiple labels

Arguments

• message - Commit message to classify
• top_n - Maximum number of categories to return
• min_confidence - Minimum confidence threshold (0.0 to 1.0)

Returns

• Ok(MultiLabelClassification) with top-N categories

Examples

use organizational_intelligence_plugin::classifier::HybridClassifier;

let classifier = HybridClassifier::new_rule_based();
let result = classifier.classify_multi_label("fix: null pointer in parser", 3, 0.60).unwrap();
println!("Primary: {:?} ({:.2})", result.primary_category, result.primary_confidence);

Trait Implementations

impl Default for HybridClassifier

fn default() -> Self

Returns the “default value” for a type.