oscura 0.0.1__py3-none-any.whl → 0.1.1__py3-none-any.whl

oscura/quality/scoring.py

@@ -0,0 +1,616 @@
+"""Analysis quality scoring for Oscura.
+
+This module provides quality scoring and reliability categorization for
+analysis results, enabling users to assess confidence in automated findings.
+
+
+Example:
+    >>> from oscura.quality.scoring import AnalysisQualityScore, ReliabilityCategory
+    >>> score = AnalysisQualityScore(
+    ...     confidence=0.85,
+    ...     category=ReliabilityCategory.HIGH,
+    ...     data_quality_factor=0.9,
+    ...     sample_sufficiency=0.8,
+    ...     method_reliability=0.85,
+    ... )
+    >>> print(score.explain())
+    >>> recommendations = score.get_recommendations()
+
+References:
+    - Quality scoring for automated analysis results
+"""
+
+from __future__ import annotations
+
+import logging
+from dataclasses import dataclass, field
+from enum import Enum
+from typing import TYPE_CHECKING, Any
+
+import numpy as np
+
+if TYPE_CHECKING:
+    from numpy.typing import NDArray
+
+logger = logging.getLogger(__name__)
+
+
+class ReliabilityCategory(Enum):
+    """Reliability categories for analysis results.
+
+    Attributes:
+        HIGH: Result is highly reliable (confidence >= 0.8)
+        MEDIUM: Result has moderate reliability (0.6 <= confidence < 0.8)
+        LOW: Result has low reliability (0.4 <= confidence < 0.6)
+        UNRELIABLE: Result is unreliable (confidence < 0.4)
+    """
+
+    HIGH = "high"
+    MEDIUM = "medium"
+    LOW = "low"
+    UNRELIABLE = "unreliable"
+
+    @classmethod
+    def from_confidence(cls, confidence: float) -> ReliabilityCategory:
+        """Get category from confidence score.
+
+        Args:
+            confidence: Confidence value in range [0, 1]
+
+        Returns:
+            Appropriate ReliabilityCategory
+        """
+        if confidence >= 0.8:
+            return cls.HIGH
+        elif confidence >= 0.6:
+            return cls.MEDIUM
+        elif confidence >= 0.4:
+            return cls.LOW
+        else:
+            return cls.UNRELIABLE
+
+
+@dataclass
+class AnalysisQualityScore:
+    """Quality score for an analysis result.
+
+    Attributes:
+        confidence: Overall confidence in result (0-1)
+        category: Reliability category
+        data_quality_factor: Quality of input data (0-1)
+        sample_sufficiency: Sufficiency of sample count (0-1)
+        method_reliability: Inherent reliability of method (0-1)
+        factors: Additional contributing factors
+        warnings: Quality warnings
+        metadata: Additional metadata
+
+    Example:
+        >>> score = AnalysisQualityScore(
+        ...     confidence=0.85,
+        ...     category=ReliabilityCategory.HIGH,
+        ...     data_quality_factor=0.9,
+        ...     sample_sufficiency=0.8,
+        ...     method_reliability=0.85,
+        ... )
+        >>> if score.is_reliable:
+        ...     print("Result is reliable")
+    """
+
+    confidence: float
+    category: ReliabilityCategory
+    data_quality_factor: float
+    sample_sufficiency: float
+    method_reliability: float
+    factors: dict[str, float] = field(default_factory=dict)
+    warnings: list[str] = field(default_factory=list)
+    metadata: dict[str, Any] = field(default_factory=dict)
+
+    def __post_init__(self) -> None:
+        """Validate score values."""
+        # Ensure confidence is in valid range
+        if not 0 <= self.confidence <= 1:
+            raise ValueError(f"Confidence must be in [0, 1], got {self.confidence}")
+
+        # Validate factors
+        for name, value in [
+            ("data_quality_factor", self.data_quality_factor),
+            ("sample_sufficiency", self.sample_sufficiency),
+            ("method_reliability", self.method_reliability),
+        ]:
+            if not 0 <= value <= 1:
+                raise ValueError(f"{name} must be in [0, 1], got {value}")
+
+    @property
+    def is_reliable(self) -> bool:
+        """Check if result is reliable (medium confidence or higher).
+
+        Returns:
+            True if category is HIGH or MEDIUM
+        """
+        return self.category in (ReliabilityCategory.HIGH, ReliabilityCategory.MEDIUM)
+
+    def explain(self, include_factors: bool = True) -> str:
+        """Generate human-readable explanation of the quality score.
+
+        Args:
+            include_factors: Whether to include factor breakdown
+
+        Returns:
+            Human-readable explanation string
+
+        Example:
+            >>> print(score.explain())
+            ✓ High confidence result (85.0%)
+
+            Contributing factors:
+              ✓ Data Quality Factor: 90.0%
+              ✓ Sample Sufficiency: 80.0%
+              ✓ Method Reliability: 85.0%
+        """
+        lines = []
+
+        # Overall assessment
+        if self.category == ReliabilityCategory.HIGH:
+            lines.append(f"✓ High confidence result ({self.confidence:.1%})")
+        elif self.category == ReliabilityCategory.MEDIUM:
+            lines.append(f"◐ Medium confidence result ({self.confidence:.1%})")
+        elif self.category == ReliabilityCategory.LOW:
+            lines.append(f"◯ Low confidence result ({self.confidence:.1%})")
+        else:
+            lines.append(f"✗ Unreliable result ({self.confidence:.1%})")
+
+        # Factor breakdown
+        if include_factors and self.factors:
+            lines.append("\nContributing factors:")
+            for factor_name, factor_value in sorted(self.factors.items()):
+                status = "✓" if factor_value >= 0.7 else "◐" if factor_value >= 0.4 else "✗"
+                lines.append(
+                    f"  {status} {factor_name.replace('_', ' ').title()}: {factor_value:.1%}"
+                )
+
+        # Warnings
+        if self.warnings:
+            lines.append("\nWarnings:")
+            for warning in self.warnings:
+                lines.append(f"  ⚠ {warning}")
+
+        return "\n".join(lines)
+
+    def get_recommendations(self) -> list[str]:
+        """Get actionable recommendations to improve result quality.
+
+        Returns:
+            List of recommendation strings
+
+        Example:
+            >>> recommendations = score.get_recommendations()
+            >>> for rec in recommendations:
+            ...     print(rec)
+            Consider improving input signal quality (filtering, averaging)
+        """
+        recommendations = []
+
+        if self.data_quality_factor < 0.5:
+            recommendations.append("Consider improving input signal quality (filtering, averaging)")
+
+        if self.sample_sufficiency < 0.5:
+            recommendations.append("Capture more data points for reliable analysis")
+
+        if "snr" in str(self.warnings).lower():
+            recommendations.append("Use a bandpass filter to improve SNR")
+
+        if "clipping" in str(self.warnings).lower():
+            recommendations.append("Adjust input gain to avoid signal clipping")
+
+        if not recommendations:
+            recommendations.append("Result quality is acceptable")
+
+        return recommendations
+
+    def to_dict(self) -> dict[str, Any]:
+        """Convert to dictionary for serialization.
+
+        Returns:
+            Dictionary representation of quality score.
+        """
+        return {
+            "confidence": self.confidence,
+            "category": self.category.value,
+            "is_reliable": self.is_reliable,
+            "data_quality_factor": self.data_quality_factor,
+            "method_reliability": self.method_reliability,
+            "sample_sufficiency": self.sample_sufficiency,
+            "factors": self.factors,
+            "warnings": self.warnings,
+            "metadata": self.metadata,
+        }
+
+
+def calculate_quality_score(
+    data_quality_factor: float,
+    sample_sufficiency: float,
+    method_reliability: float,
+    *,
+    weights: tuple[float, float, float] | None = None,
+    warnings: list[str] | None = None,
+    factors: dict[str, float] | None = None,
+    metadata: dict[str, Any] | None = None,
+) -> AnalysisQualityScore:
+    """Calculate overall quality score from component factors.
+
+    Args:
+        data_quality_factor: Quality of input data (0-1)
+        sample_sufficiency: Sufficiency of sample count (0-1)
+        method_reliability: Inherent reliability of method (0-1)
+        weights: Optional custom weights (data, sample, method), defaults to (0.4, 0.3, 0.3)
+        warnings: Optional quality warnings
+        factors: Optional additional factors
+        metadata: Optional metadata
+
+    Returns:
+        AnalysisQualityScore with computed confidence
+
+    Example:
+        >>> score = calculate_quality_score(
+        ...     data_quality_factor=0.9,
+        ...     sample_sufficiency=0.8,
+        ...     method_reliability=0.85,
+        ... )
+        >>> print(f"Confidence: {score.confidence:.1%}")
+    """
+    if weights is None:
+        weights = (0.4, 0.3, 0.3)
+
+    w_data, w_sample, w_method = weights
+
+    # Calculate weighted confidence
+    confidence = (
+        w_data * data_quality_factor + w_sample * sample_sufficiency + w_method * method_reliability
+    )
+
+    # Determine category
+    category = ReliabilityCategory.from_confidence(confidence)
+
+    # Build factors dictionary
+    all_factors = {
+        "data_quality_factor": data_quality_factor,
+        "sample_sufficiency": sample_sufficiency,
+        "method_reliability": method_reliability,
+    }
+    if factors:
+        all_factors.update(factors)
+
+    return AnalysisQualityScore(
+        confidence=confidence,
+        category=category,
+        data_quality_factor=data_quality_factor,
+        sample_sufficiency=sample_sufficiency,
+        method_reliability=method_reliability,
+        factors=all_factors,
+        warnings=warnings or [],
+        metadata=metadata or {},
+    )
+
+
+@dataclass
+class DataQualityMetrics:
+    """Metrics describing input data quality.
+
+        QUAL-002: Data quality assessment
+
+    Attributes:
+        snr_db: Signal-to-noise ratio in decibels
+        sample_count: Number of samples in the data
+        has_clipping: Whether the signal shows clipping
+        has_saturation: Whether the signal shows saturation
+        noise_floor: Estimated noise floor level
+        completeness: Fraction of non-NaN values (0-1)
+    """
+
+    snr_db: float | None = None
+    sample_count: int = 0
+    has_clipping: bool = False
+    has_saturation: bool = False
+    noise_floor: float | None = None
+    completeness: float = 1.0  # Fraction of non-NaN values
+
+    def to_factor(self) -> float:
+        """Convert metrics to single quality factor (0-1).
+
+        Returns:
+            Quality factor between 0 and 1.
+        """
+        factors = []
+
+        # SNR contribution
+        if self.snr_db is not None:
+            snr_factor = min(1.0, max(0.0, self.snr_db / 40.0))
+            factors.append(snr_factor)
+
+        # Sample count contribution (diminishing returns after 1000)
+        sample_factor = min(1.0, np.log10(max(1, self.sample_count)) / 4.0)
+        factors.append(sample_factor)
+
+        # Clipping/saturation penalties
+        if self.has_clipping:
+            factors.append(0.7)
+        if self.has_saturation:
+            factors.append(0.6)
+
+        # Completeness
+        factors.append(self.completeness)
+
+        return float(np.mean(factors)) if factors else 0.5
+
+
+# Method reliability scores (based on algorithm characteristics)
+#: Method reliability tracking
+METHOD_RELIABILITY: dict[str, float] = {
+    # High reliability methods
+    "fft": 0.95,
+    "welch": 0.90,
+    "autocorrelation": 0.85,
+    "histogram": 0.95,
+    "statistics": 0.95,
+    # Medium reliability methods
+    "edge_detection": 0.80,
+    "zero_crossing": 0.75,
+    "peak_detection": 0.70,
+    "pattern_matching": 0.75,
+    # Lower reliability methods (heuristic-based)
+    "protocol_inference": 0.60,
+    "signal_classification": 0.65,
+    "anomaly_detection": 0.60,
+}
+
+
+def assess_data_quality(
+    data: NDArray[np.float64], sample_rate: float | None = None
+) -> DataQualityMetrics:
+    """Assess quality of input data.
+
+        QUAL-002: Data quality assessment
+
+    Args:
+        data: Input data array
+        sample_rate: Sample rate in Hz (optional)
+
+    Returns:
+        DataQualityMetrics with quality assessment
+    """
+    metrics = DataQualityMetrics()
+
+    try:
+        # Sample count
+        metrics.sample_count = len(data)
+
+        # Check for NaN/Inf
+        valid_mask = np.isfinite(data)
+        metrics.completeness = float(np.mean(valid_mask))
+
+        if metrics.completeness < 0.01:
+            return metrics
+
+        valid_data = data[valid_mask]
+
+        # Check for clipping (values at min/max bounds)
+        data_range = np.ptp(valid_data)
+        if data_range > 0:
+            min_count = np.sum(valid_data == np.min(valid_data))
+            max_count = np.sum(valid_data == np.max(valid_data))
+            clip_threshold = 0.01 * len(valid_data)
+            metrics.has_clipping = min_count > clip_threshold or max_count > clip_threshold
+
+        # Estimate SNR using signal variance vs noise floor
+        # Use median absolute deviation for robust noise estimation
+        median = np.median(valid_data)
+        mad = np.median(np.abs(valid_data - median)) * 1.4826
+        metrics.noise_floor = float(mad)
+
+        signal_power = float(np.var(valid_data))
+        noise_power = mad**2
+
+        if noise_power > 0:
+            snr_linear = signal_power / noise_power
+            metrics.snr_db = float(10 * np.log10(max(1e-10, snr_linear)))
+
+    except Exception as e:
+        logger.debug(f"Error assessing data quality: {e}")
+
+    return metrics
+
+
+def score_analysis_result(
+    result: Any,
+    method_name: str,
+    data: NDArray[np.float64] | None = None,
+    data_quality: DataQualityMetrics | None = None,
+    min_samples: int = 10,
+) -> AnalysisQualityScore:
+    """Score the quality of an analysis result.
+
+        QUAL-001: Quality scoring foundation
+
+    Args:
+        result: The analysis result to score
+        method_name: Name of the analysis method
+        data: Input data (for quality assessment)
+        data_quality: Pre-computed data quality metrics
+        min_samples: Minimum samples for reliable result
+
+    Returns:
+        AnalysisQualityScore with confidence and factors
+    """
+    factors = {}
+    warnings = []
+
+    # Get data quality
+    if data_quality is None and data is not None:
+        data_quality = assess_data_quality(data)
+
+    # Data quality factor
+    if data_quality is not None:
+        data_factor = data_quality.to_factor()
+        factors["data_quality"] = data_factor
+
+        if data_quality.has_clipping:
+            warnings.append("Input data shows clipping")
+        if data_quality.snr_db is not None and data_quality.snr_db < 20:
+            warnings.append(f"Low SNR ({data_quality.snr_db:.1f} dB)")
+    else:
+        data_factor = 0.5
+        factors["data_quality"] = data_factor
+
+    # Method reliability
+    method_key = method_name.lower().split(".")[-1].replace("_", "")
+    method_reliability = METHOD_RELIABILITY.get(method_key, 0.7)
+
+    # Check for partial matches
+    for key, reliability in METHOD_RELIABILITY.items():
+        if key in method_name.lower():
+            method_reliability = reliability
+            break
+
+    factors["method_reliability"] = method_reliability
+
+    # Sample sufficiency
+    if data_quality is not None:
+        sample_sufficiency = min(1.0, data_quality.sample_count / (min_samples * 10))
+        if data_quality.sample_count < min_samples:
+            warnings.append(f"Insufficient samples ({data_quality.sample_count} < {min_samples})")
+    else:
+        sample_sufficiency = 0.5
+    factors["sample_sufficiency"] = sample_sufficiency
+
+    # Result-specific scoring
+    result_factor = _score_result_value(result)
+    factors["result_validity"] = result_factor
+
+    # Combine factors
+    confidence = (
+        data_factor * 0.3
+        + method_reliability * 0.25
+        + sample_sufficiency * 0.25
+        + result_factor * 0.2
+    )
+
+    # Determine category from confidence
+    category = ReliabilityCategory.from_confidence(confidence)
+
+    return AnalysisQualityScore(
+        confidence=confidence,
+        category=category,
+        data_quality_factor=data_factor,
+        method_reliability=method_reliability,
+        sample_sufficiency=sample_sufficiency,
+        factors=factors,
+        warnings=warnings,
+    )
+
+
+def _score_result_value(result: Any) -> float:
+    """Score result validity based on value characteristics.
+
+    Args:
+        result: Analysis result to score.
+
+    Returns:
+        Validity score between 0 and 1.
+    """
+    if result is None:
+        return 0.0
+
+    # Handle numeric results
+    if isinstance(result, int | float):
+        if np.isnan(result) or np.isinf(result):
+            return 0.0
+        return 1.0
+
+    # Handle array results
+    if isinstance(result, np.ndarray):
+        valid_ratio = np.mean(np.isfinite(result))
+        return float(valid_ratio)
+
+    # Handle dict results
+    if isinstance(result, dict):
+        if not result:
+            return 0.3
+        return 1.0
+
+    # Handle list results
+    if isinstance(result, list):
+        if not result:
+            return 0.3
+        return 1.0
+
+    return 0.7  # Default for other types
+
+
+def combine_quality_scores(
+    scores: list[AnalysisQualityScore],
+    weights: list[float] | None = None,
+) -> AnalysisQualityScore:
+    """Combine multiple quality scores into one.
+
+    Args:
+        scores: List of quality scores to combine
+        weights: Optional weights for each score
+
+    Returns:
+        Combined quality score
+    """
+    if not scores:
+        return AnalysisQualityScore(
+            confidence=0.0,
+            category=ReliabilityCategory.UNRELIABLE,
+            data_quality_factor=0.0,
+            method_reliability=0.0,
+            sample_sufficiency=0.0,
+        )
+
+    if weights is None:
+        weights = [1.0] * len(scores)
+
+    total_weight = sum(weights)
+
+    combined_confidence = (
+        sum(s.confidence * w for s, w in zip(scores, weights, strict=True)) / total_weight
+    )
+    combined_data = (
+        sum(s.data_quality_factor * w for s, w in zip(scores, weights, strict=True)) / total_weight
+    )
+    combined_method = (
+        sum(s.method_reliability * w for s, w in zip(scores, weights, strict=True)) / total_weight
+    )
+    combined_samples = (
+        sum(s.sample_sufficiency * w for s, w in zip(scores, weights, strict=True)) / total_weight
+    )
+
+    # Aggregate warnings
+    all_warnings = []
+    for score in scores:
+        all_warnings.extend(score.warnings)
+
+    # Determine category
+    category = ReliabilityCategory.from_confidence(combined_confidence)
+
+    return AnalysisQualityScore(
+        confidence=combined_confidence,
+        category=category,
+        data_quality_factor=combined_data,
+        method_reliability=combined_method,
+        sample_sufficiency=combined_samples,
+        warnings=list(set(all_warnings)),
+    )
+
+
+__all__ = [
+    "METHOD_RELIABILITY",
+    "AnalysisQualityScore",
+    "DataQualityMetrics",
+    "ReliabilityCategory",
+    "assess_data_quality",
+    "calculate_quality_score",
+    "combine_quality_scores",
+    "score_analysis_result",
+]