Struct TrainingDataExtractor 

pub struct TrainingDataExtractor { /* private fields */ }

Training data extractor

Implementations

impl TrainingDataExtractor

pub fn new(min_confidence: f32) -> Self

Create a new training data extractor

Arguments

• min_confidence - Minimum confidence threshold for auto-labeling (0.6-0.9)

Examples

use organizational_intelligence_plugin::training::TrainingDataExtractor;

let extractor = TrainingDataExtractor::new(0.75);

pub fn extract_training_data( &self, commits: &[CommitInfo], _repository_name: &str, ) -> Result<Vec<TrainingExample>>

Extract training examples from commit history

Filters commits and auto-labels using rule-based classifier.

Arguments

• commits - Raw commit history
• repository_name - Name of the repository

Returns

• Ok(Vec<TrainingExample>) - Labeled training examples
• Err - If extraction fails

Examples

use organizational_intelligence_plugin::training::TrainingDataExtractor;
use organizational_intelligence_plugin::git::CommitInfo;

let extractor = TrainingDataExtractor::new(0.75);
let commits = vec![
    CommitInfo {
        hash: "abc123".to_string(),
        message: "fix: null pointer dereference".to_string(),
        author: "dev@example.com".to_string(),
        timestamp: 1234567890,
        files_changed: 2,
        lines_added: 10,
        lines_removed: 5,
    },
];

let examples = extractor.extract_training_data(&commits, "test-repo").unwrap();
assert_eq!(examples.len(), 1);

pub fn create_splits( &self, examples: &[TrainingExample], repositories: &[String], ) -> Result<TrainingDataset>

Create train/test/validation splits

Uses 70/15/15 split (train/validation/test) as recommended by the spec.

Arguments

• examples - Labeled training examples
• repositories - List of repository names

Returns

• Ok(TrainingDataset) - Dataset with splits
• Err - If split fails

Examples

use organizational_intelligence_plugin::training::TrainingDataExtractor;
use organizational_intelligence_plugin::training::TrainingExample;
use organizational_intelligence_plugin::classifier::DefectCategory;

let extractor = TrainingDataExtractor::new(0.75);
let examples = vec![
    TrainingExample {
        message: "fix: bug".to_string(),
        label: DefectCategory::MemorySafety,
        confidence: 0.85,
        commit_hash: "abc".to_string(),
        author: "dev".to_string(),
        timestamp: 123,
        lines_added: 5,
        lines_removed: 2,
        files_changed: 1,
        error_code: None,
        clippy_lint: None,
        has_suggestion: false,
        suggestion_applicability: None,
        source: organizational_intelligence_plugin::citl::TrainingSource::CommitMessage,
    },
];

let dataset = extractor.create_splits(&examples, &["repo1".to_string()]).unwrap();
assert!(dataset.train.len() + dataset.validation.len() + dataset.test.len() == 1);

pub fn get_statistics(&self, examples: &[TrainingExample]) -> String

Get statistics about extracted training data

Arguments

• examples - Training examples

Returns

• Formatted statistics string

Trait Implementations

impl Default for TrainingDataExtractor

fn default() -> Self

Returns the “default value” for a type.