File size: 8,114 Bytes
a3934b1
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
224
225
226
227
228
229
230
231
# verification_metrics.py
"""
Verification metrics calculator service.

Provides methods for calculating accuracy, confusion matrices, and error patterns
from verification records.
"""

from typing import Dict, List, Any
from src.core.verification_models import VerificationRecord


class VerificationMetricsCalculator:
    """Calculates performance metrics from verification records."""

    @staticmethod
    def calculate_accuracy(records: List[VerificationRecord]) -> float:
        """
        Calculate overall accuracy from verification records.
        
        Accuracy = (correct_count / total_count) * 100
        
        Args:
            records: List of verification records
            
        Returns:
            Accuracy as a percentage (0-100), or 0 if no records
        """
        if not records:
            return 0.0
        
        correct_count = sum(1 for record in records if record.is_correct)
        return (correct_count / len(records)) * 100

    @staticmethod
    def calculate_accuracy_by_type(
        records: List[VerificationRecord],
    ) -> Dict[str, float]:
        """
        Calculate accuracy for each classification type.
        
        For each type (green, yellow, red), calculates:
        accuracy = (correct_count_for_type / total_count_for_type) * 100
        
        Args:
            records: List of verification records
            
        Returns:
            Dictionary with keys "green", "yellow", "red" and accuracy percentages
        """
        accuracy_by_type = {}
        
        for classification_type in ["green", "yellow", "red"]:
            type_records = [
                r for r in records
                if r.classifier_decision == classification_type
            ]
            
            if type_records:
                correct_count = sum(1 for r in type_records if r.is_correct)
                accuracy_by_type[classification_type] = (
                    correct_count / len(type_records) * 100
                )
            else:
                accuracy_by_type[classification_type] = 0.0
        
        return accuracy_by_type

    @staticmethod
    def calculate_confusion_matrix(
        records: List[VerificationRecord],
    ) -> Dict[str, Dict[str, int]]:
        """
        Generate a confusion matrix from verification records.
        
        The confusion matrix shows:
        - Rows: classifier decisions (what the classifier said)
        - Columns: ground truth labels (what the verifier said)
        - Values: count of records in each cell
        
        Args:
            records: List of verification records
            
        Returns:
            Dictionary with structure:
            {
                "green": {"green": count, "yellow": count, "red": count},
                "yellow": {"green": count, "yellow": count, "red": count},
                "red": {"green": count, "yellow": count, "red": count},
            }
        """
        # Initialize matrix with zeros
        matrix = {
            "green": {"green": 0, "yellow": 0, "red": 0},
            "yellow": {"green": 0, "yellow": 0, "red": 0},
            "red": {"green": 0, "yellow": 0, "red": 0},
        }
        
        # Populate matrix
        for record in records:
            classifier_decision = record.classifier_decision
            ground_truth = record.ground_truth_label
            matrix[classifier_decision][ground_truth] += 1
        
        return matrix

    @staticmethod
    def generate_error_patterns(
        records: List[VerificationRecord],
    ) -> List[str]:
        """
        Detect common error patterns from verification records.
        
        Identifies patterns like:
        - "Often misclassifies YELLOW as GREEN"
        - "Frequently misses RED indicators"
        
        Args:
            records: List of verification records
            
        Returns:
            List of error pattern descriptions
        """
        if not records:
            return []
        
        patterns = []
        
        # Get confusion matrix
        matrix = VerificationMetricsCalculator.calculate_confusion_matrix(records)
        
        # Analyze each classification type
        for classifier_type in ["green", "yellow", "red"]:
            type_records = [
                r for r in records
                if r.classifier_decision == classifier_type
            ]
            
            if not type_records:
                continue
            
            # Find most common misclassification
            misclassifications = {}
            for record in type_records:
                if not record.is_correct:
                    ground_truth = record.ground_truth_label
                    misclassifications[ground_truth] = (
                        misclassifications.get(ground_truth, 0) + 1
                    )
            
            if misclassifications:
                most_common_wrong = max(
                    misclassifications.items(), key=lambda x: x[1]
                )
                wrong_type, wrong_count = most_common_wrong
                
                # Calculate percentage of misclassifications
                error_rate = (wrong_count / len(type_records)) * 100
                
                if error_rate >= 20:  # Only report if >= 20% error rate
                    pattern = (
                        f"Often misclassifies {classifier_type.upper()} "
                        f"as {wrong_type.upper()} ({error_rate:.0f}% of {classifier_type.upper()} cases)"
                    )
                    patterns.append(pattern)
        
        # Analyze missed classifications (false negatives)
        for ground_truth_type in ["green", "yellow", "red"]:
            # Find records where classifier missed this type
            missed = [
                r for r in records
                if r.ground_truth_label == ground_truth_type
                and r.classifier_decision != ground_truth_type
            ]
            
            if missed:
                missed_rate = (len(missed) / len(records)) * 100
                
                if missed_rate >= 10:  # Only report if >= 10% miss rate
                    pattern = (
                        f"Frequently misses {ground_truth_type.upper()} indicators "
                        f"({missed_rate:.0f}% of all messages)"
                    )
                    patterns.append(pattern)
        
        return patterns

    @staticmethod
    def get_metrics_summary(records: List[VerificationRecord]) -> Dict[str, Any]:
        """
        Get a comprehensive summary of all metrics.
        
        Args:
            records: List of verification records
            
        Returns:
            Dictionary containing all calculated metrics
        """
        if not records:
            return {
                "total_records": 0,
                "correct_count": 0,
                "incorrect_count": 0,
                "accuracy": 0.0,
                "accuracy_by_type": {"green": 0.0, "yellow": 0.0, "red": 0.0},
                "confusion_matrix": {
                    "green": {"green": 0, "yellow": 0, "red": 0},
                    "yellow": {"green": 0, "yellow": 0, "red": 0},
                    "red": {"green": 0, "yellow": 0, "red": 0},
                },
                "error_patterns": [],
            }
        
        correct_count = sum(1 for r in records if r.is_correct)
        
        return {
            "total_records": len(records),
            "correct_count": correct_count,
            "incorrect_count": len(records) - correct_count,
            "accuracy": VerificationMetricsCalculator.calculate_accuracy(records),
            "accuracy_by_type": (
                VerificationMetricsCalculator.calculate_accuracy_by_type(records)
            ),
            "confusion_matrix": (
                VerificationMetricsCalculator.calculate_confusion_matrix(records)
            ),
            "error_patterns": (
                VerificationMetricsCalculator.generate_error_patterns(records)
            ),
        }