spawn_access_control/
model_explainer.rs

use crate::model_optimizer::ModelParameters;
use crate::ml_metrics::ModelMetrics;
use serde::Serialize;
use std::collections::HashMap;

#[derive(Debug, Serialize)]
pub struct ModelExplanation {
    pub feature_importance: HashMap<String, f64>,
    pub decision_paths: Vec<DecisionPath>,
    pub performance_metrics: ExplainedMetrics,
    pub security_impact: SecurityImpactAnalysis,
}

#[derive(Debug, Serialize)]
pub struct DecisionPath {
    pub path: Vec<DecisionNode>,
    pub confidence: f64,
    pub impact: f64,
}

#[derive(Debug, Serialize)]
pub struct DecisionNode {
    pub feature: String,
    pub threshold: f64,
    pub direction: String,
}

#[derive(Debug, Serialize)]
pub struct ExplainedMetrics {
    pub metrics: ModelMetrics,
    pub explanations: HashMap<String, String>,
    pub recommendations: Vec<String>,
}

#[derive(Debug, Clone, Serialize)]
pub struct SecurityImpactAnalysis {
    pub false_positive_impact: f64,
    pub risk_factors: Vec<RiskFactor>,
}

impl SecurityImpactAnalysis {
    pub fn analyze_security_impact(&self, metrics: &ModelMetrics) -> Self {
        Self {
            false_positive_impact: 1.0 - metrics.precision,
            risk_factors: self.analyze_risk_factors(metrics),
        }
    }

    fn analyze_risk_factors(&self, metrics: &ModelMetrics) -> Vec<RiskFactor> {
        let mut factors = Vec::new();
        
        if metrics.precision < 0.9 {
            factors.push(RiskFactor {
                name: "High False Positive Rate".to_string(),
                impact_score: 1.0 - metrics.precision,
            });
        }
        
        factors
    }
}

#[derive(Clone, Debug, Serialize)]
pub struct RiskFactor {
    pub name: String,
    pub impact_score: f64,
}

pub struct ModelExplainer {
    feature_names: Vec<String>,
    #[allow(dead_code)]
    security_config: SecurityConfig,
}

#[derive(Clone)]
pub struct SecurityConfig {
    pub false_positive_weight: f64,
    pub false_negative_weight: f64,
    pub risk_threshold: f64,
}

impl ModelExplainer {
    pub fn new(feature_names: Vec<String>, security_config: SecurityConfig) -> Self {
        Self {
            feature_names,
            security_config,
        }
    }

    pub fn explain_model(&self, params: &ModelParameters, metrics: &ModelMetrics) -> ModelExplanation {
        ModelExplanation {
            feature_importance: self.calculate_feature_importance(params),
            decision_paths: self.analyze_decision_paths(),
            performance_metrics: self.explain_metrics(metrics),
            security_impact: self.analyze_security_impact(metrics),
        }
    }

    fn calculate_feature_importance(&self, params: &ModelParameters) -> HashMap<String, f64> {
        let mut importance = HashMap::new();
        
        // Feature önem skorlarını hesapla
        for (idx, feature) in self.feature_names.iter().enumerate() {
            let score = self.calculate_feature_score(idx, params);
            importance.insert(feature.clone(), score);
        }

        // Skorları normalize et
        let total: f64 = importance.values().sum();
        for score in importance.values_mut() {
            *score /= total;
        }

        importance
    }

    fn calculate_feature_score(&self, feature_idx: usize, params: &ModelParameters) -> f64 {
        // Feature önem skoru hesaplama mantığı
        let base_score = 1.0 / self.feature_names.len() as f64;
        let depth_factor = (-((params.max_depth as f64 - 10.0).powi(2)) / 100.0).exp();
        
        base_score * depth_factor * (1.0 + (feature_idx as f64 / 10.0))
    }

    fn analyze_decision_paths(&self) -> Vec<DecisionPath> {
        // Örnek karar yolları analizi
        vec![
            DecisionPath {
                path: vec![
                    DecisionNode {
                        feature: "time_of_day".to_string(),
                        threshold: 18.0,
                        direction: "greater_than".to_string(),
                    },
                    DecisionNode {
                        feature: "failed_attempts".to_string(),
                        threshold: 3.0,
                        direction: "less_than".to_string(),
                    },
                ],
                confidence: 0.85,
                impact: 0.7,
            }
        ]
    }

    fn explain_metrics(&self, metrics: &ModelMetrics) -> ExplainedMetrics {
        let mut explanations = HashMap::new();
        let mut recommendations = Vec::new();

        // Metrikleri açıkla
        explanations.insert(
            "f1_score".to_string(),
            format!("F1 score of {:.2} indicates balanced performance", metrics.f1_score)
        );

        // Öneriler oluştur
        if metrics.f1_score < 0.8 {
            recommendations.push("Consider increasing training data diversity".to_string());
        }
        if metrics.precision < metrics.recall {
            recommendations.push("Model might be too aggressive, consider adjusting threshold".to_string());
        }

        ExplainedMetrics {
            metrics: metrics.clone(),
            explanations,
            recommendations,
        }
    }

    fn analyze_security_impact(&self, metrics: &ModelMetrics) -> SecurityImpactAnalysis {
        SecurityImpactAnalysis {
            false_positive_impact: 1.0 - metrics.precision,
            risk_factors: vec![], // Boş risk faktörleri ile başla
        }
    }
}
spawn_access_control/model_explainer.rs

spawn_access_control/
model_explainer.rs
