evaluation_module.py

# =====================================================================
# EVALUATION MODULE - Untuk Pengerjaan Revisi Karya Tulis
# =====================================================================
# Module ini menyediakan tools untuk:
# 1. Experiment 1: Topic Classification Accuracy (LDA)
# 2. Experiment 2: Speed Improvement (manual vs system)
# 3. Experiment 3: Deduplication Accuracy
# 4. Experiment 4: Inter-Model Consistency (OpenAI vs Gemini)
# 5. Experiment 5: Scalability Testing (concurrent users)
# 6. Experiment 6: User Satisfaction Survey
# =====================================================================

import pandas as pd
import numpy as np
import json
import time
import requests
from datetime import datetime
from collections import Counter
from sklearn.metrics import precision_score, recall_score, f1_score, confusion_matrix
from sklearn.decomposition import LatentDirichletAllocation
from sklearn.feature_extraction.text import CountVectorizer
import matplotlib.pyplot as plt
import seaborn as sns
from tabulate import tabulate

# =====================================================================
# EXPERIMENT 1: TOPIC CLASSIFICATION ACCURACY (LDA)
# =====================================================================

class TopicClassificationEvaluation:
    """
    Evaluate LDA topic classification accuracy
    
    Example:
        eval = TopicClassificationEvaluation()
        results = eval.run_experiment(
            df=articles_df,
            ground_truth=manual_labels,  # List of (article_id, true_topic_label)
            n_topics=5
        )
    """
    
    def __init__(self):
        self.lda_model = None
        self.vectorizer = None
        
    def prepare_data(self, df, text_column='full_text'):
        """Prepare text data for LDA"""
        texts = df[text_column].fillna('').astype(str)
        self.vectorizer = CountVectorizer(
            max_df=0.9, 
            min_df=2, 
            max_features=1000,
            stop_words='english'
        )
        return self.vectorizer.fit_transform(texts)
    
    def train_lda(self, doc_term_matrix, n_topics=5):
        """Train LDA model"""
        self.lda_model = LatentDirichletAllocation(
            n_components=n_topics,
            random_state=42,
            max_iter=20
        )
        self.lda_model.fit(doc_term_matrix)
        return self.lda_model
    
    def get_topic_assignments(self, doc_term_matrix):
        """Get topic assignment for each document"""
        return self.lda_model.transform(doc_term_matrix).argmax(axis=1)
    
    def run_experiment(self, df, ground_truth_labels, n_topics=5):
        """
        Run full evaluation experiment
        
        Args:
            df: DataFrame with articles
            ground_truth_labels: List of actual topic labels (same order as df)
            n_topics: Number of topics for LDA
            
        Returns:
            dict: Evaluation metrics
        """
        print(f"[EXP-1] Training LDA with {n_topics} topics...")
        
        # Prepare & train
        doc_term_matrix = self.prepare_data(df)
        self.train_lda(doc_term_matrix, n_topics)
        
        # Get predictions
        predicted_topics = self.get_topic_assignments(doc_term_matrix)
        
        # Map to ground truth (assuming ground_truth_labels are 0-4)
        y_true = np.array(ground_truth_labels)
        y_pred = predicted_topics
        
        # Calculate metrics
        precision = precision_score(y_true, y_pred, average='weighted', zero_division=0)
        recall = recall_score(y_true, y_pred, average='weighted', zero_division=0)
        f1 = f1_score(y_true, y_pred, average='weighted', zero_division=0)
        
        # Confusion matrix
        cm = confusion_matrix(y_true, y_pred)
        
        # Per-class metrics
        precision_per_class = precision_score(y_true, y_pred, average=None, zero_division=0)
        recall_per_class = recall_score(y_true, y_pred, average=None, zero_division=0)
        
        results = {
            'precision': precision,
            'recall': recall,
            'f1_score': f1,
            'precision_per_class': precision_per_class.tolist(),
            'recall_per_class': recall_per_class.tolist(),
            'confusion_matrix': cm.tolist(),
            'n_topics': n_topics,
            'n_samples': len(df),
            'timestamp': datetime.now().isoformat()
        }
        
        print(f"[EXP-1] Results: Precision={precision:.3f}, Recall={recall:.3f}, F1={f1:.3f}")
        return results
    
    def print_report(self, results):
        """Print evaluation report"""
        print("\n" + "="*70)
        print("EXPERIMENT 1: TOPIC CLASSIFICATION ACCURACY (LDA)")
        print("="*70)
        
        data = [
            ["Precision", f"{results['precision']:.3f}"],
            ["Recall", f"{results['recall']:.3f}"],
            ["F1-Score", f"{results['f1_score']:.3f}"],
            ["Samples", results['n_samples']],
            ["Topics", results['n_topics']]
        ]
        print(tabulate(data, headers=["Metric", "Value"], tablefmt="grid"))
        
        print("\nPer-Class Performance:")
        for i, (p, r) in enumerate(zip(results['precision_per_class'], 
                                        results['recall_per_class'])):
            print(f"  Topic {i}: Precision={p:.3f}, Recall={r:.3f}")


# =====================================================================
# EXPERIMENT 2: SPEED IMPROVEMENT
# =====================================================================

class SpeedEvaluation:
    """
    Measure speed improvement: manual review vs system
    
    Example:
        eval = SpeedEvaluation()
        results = eval.run_experiment(
            system_search_time=120,  # seconds
            manual_review_time=8*24*3600  # seconds (8 days)
        )
    """
    
    def run_experiment(self, system_search_time, manual_review_time):
        """
        Run speed comparison
        
        Args:
            system_search_time: Time for system (seconds)
            manual_review_time: Time for manual (seconds)
            
        Returns:
            dict: Speed metrics
        """
        print("[EXP-2] Measuring speed improvement...")
        
        speedup_factor = manual_review_time / system_search_time
        speedup_percentage = ((manual_review_time - system_search_time) / manual_review_time) * 100
        
        results = {
            'system_time_seconds': system_search_time,
            'manual_time_seconds': manual_review_time,
            'speedup_factor': speedup_factor,
            'speedup_percentage': speedup_percentage,
            'time_saved_seconds': manual_review_time - system_search_time,
            'system_time_formatted': self._format_time(system_search_time),
            'manual_time_formatted': self._format_time(manual_review_time),
            'timestamp': datetime.now().isoformat()
        }
        
        print(f"[EXP-2] Manual: {self._format_time(manual_review_time)}, "
              f"System: {self._format_time(system_search_time)}, "
              f"Speedup: {speedup_percentage:.1f}%")
        return results
    
    @staticmethod
    def _format_time(seconds):
        """Format seconds to human readable"""
        if seconds < 60:
            return f"{seconds:.1f}s"
        elif seconds < 3600:
            return f"{seconds/60:.1f} min"
        elif seconds < 86400:
            return f"{seconds/3600:.1f} hours"
        else:
            return f"{seconds/86400:.1f} days"
    
    def print_report(self, results):
        """Print evaluation report"""
        print("\n" + "="*70)
        print("EXPERIMENT 2: SPEED IMPROVEMENT")
        print("="*70)
        
        data = [
            ["Manual Review Time", results['manual_time_formatted']],
            ["System Processing Time", results['system_time_formatted']],
            ["Speedup Factor", f"{results['speedup_factor']:.1f}x"],
            ["Time Saved", f"{results['speedup_percentage']:.1f}%"],
        ]
        print(tabulate(data, headers=["Metric", "Value"], tablefmt="grid"))


# =====================================================================
# EXPERIMENT 3: DEDUPLICATION ACCURACY
# =====================================================================

class DeduplicationEvaluation:
    """
    Evaluate deduplication accuracy (precision & recall)
    
    Example:
        eval = DeduplicationEvaluation()
        results = eval.run_experiment(
            df=articles_with_duplicates,
            ground_truth_duplicates=[(0,1), (5,7)]  # (idx1, idx2) pairs
        )
    """
    
    def run_experiment(self, df, ground_truth_duplicates=None, manual_sample_size=100):
        """
        Run deduplication accuracy evaluation
        
        Args:
            df: DataFrame with potentially duplicate articles
            ground_truth_duplicates: List of (idx1, idx2) pairs of duplicates
            manual_sample_size: If no ground truth, simulate manual review
            
        Returns:
            dict: Deduplication metrics
        """
        print("[EXP-3] Evaluating deduplication accuracy...")
        
        # Perform deduplication
        df_dedup = self._deduplicate_by_title(df)
        duplicates_removed = len(df) - len(df_dedup)
        
        if ground_truth_duplicates is None:
            # Simulate ground truth from manual review of sample
            sample_indices = np.random.choice(len(df), min(manual_sample_size, len(df)), replace=False)
            ground_truth_duplicates = self._simulate_ground_truth(df, sample_indices)
        
        # Calculate metrics
        predictions = self._get_duplicate_predictions(df, df_dedup)
        
        tp = len([p for p in predictions if p in ground_truth_duplicates])  # True positives
        fp = len([p for p in predictions if p not in ground_truth_duplicates])  # False positives
        fn = len([g for g in ground_truth_duplicates if g not in predictions])  # False negatives
        
        precision = tp / (tp + fp) if (tp + fp) > 0 else 0
        recall = tp / (tp + fn) if (tp + fn) > 0 else 0
        f1 = 2 * (precision * recall) / (precision + recall) if (precision + recall) > 0 else 0
        
        results = {
            'original_count': len(df),
            'deduped_count': len(df_dedup),
            'duplicates_removed': duplicates_removed,
            'duplicate_percentage': (duplicates_removed / len(df)) * 100,
            'true_positives': tp,
            'false_positives': fp,
            'false_negatives': fn,
            'precision': precision,
            'recall': recall,
            'f1_score': f1,
            'timestamp': datetime.now().isoformat()
        }
        
        print(f"[EXP-3] Removed {duplicates_removed} duplicates, "
              f"Precision={precision:.3f}, Recall={recall:.3f}, F1={f1:.3f}")
        return results
    
    @staticmethod
    def _deduplicate_by_title(df):
        """Simple deduplication by title"""
        df_temp = df.copy()
        df_temp['title_lower'] = df_temp.get('title', pd.Series(dtype=str)).str.lower().str.strip()
        df_temp = df_temp.drop_duplicates(subset=['title_lower'], keep='first')
        return df_temp.drop(columns=['title_lower'])
    
    @staticmethod
    def _simulate_ground_truth(df, sample_indices):
        """Simulate ground truth duplicates"""
        # In real scenario, would have manual annotation
        # For now, simple heuristic: similar titles are likely duplicates
        duplicates = []
        for i in range(len(sample_indices)-1):
            for j in range(i+1, min(i+10, len(sample_indices))):
                idx_i = sample_indices[i]
                idx_j = sample_indices[j]
                title_i = str(df.iloc[idx_i]['title']).lower()
                title_j = str(df.iloc[idx_j]['title']).lower()
                
                # Simple similarity check
                if len(title_i) > 10 and title_i in title_j:
                    duplicates.append((idx_i, idx_j))
        
        return duplicates
    
    @staticmethod
    def _get_duplicate_predictions(df_original, df_deduped):
        """Get which rows were marked as duplicates"""
        # Find rows removed during deduplication
        removed_indices = set(df_original.index) - set(df_deduped.index)
        
        # Create pairs (original index with kept index)
        predictions = []
        for removed_idx in removed_indices:
            removed_title = str(df_original.iloc[removed_idx]['title']).lower()
            for kept_idx in df_deduped.index:
                kept_title = str(df_deduped.iloc[kept_idx]['title']).lower()
                if removed_title == kept_title:
                    predictions.append((removed_idx, kept_idx))
        
        return predictions
    
    def print_report(self, results):
        """Print evaluation report"""
        print("\n" + "="*70)
        print("EXPERIMENT 3: DEDUPLICATION ACCURACY")
        print("="*70)
        
        data = [
            ["Original Articles", results['original_count']],
            ["After Deduplication", results['deduped_count']],
            ["Duplicates Removed", results['duplicates_removed']],
            ["Duplicate %", f"{results['duplicate_percentage']:.1f}%"],
            ["Precision", f"{results['precision']:.3f}"],
            ["Recall", f"{results['recall']:.3f}"],
            ["F1-Score", f"{results['f1_score']:.3f}"]
        ]
        print(tabulate(data, headers=["Metric", "Value"], tablefmt="grid"))


# =====================================================================
# EXPERIMENT 4: INTER-MODEL CONSISTENCY
# =====================================================================

class InterModelConsistencyEvaluation:
    """
    Compare outputs from OpenAI and Gemini models
    
    Example:
        eval = InterModelConsistencyEvaluation()
        results = eval.run_experiment(
            prompts=list_of_prompts,
            openai_client=openai_client,
            gemini_api_key=gemini_key
        )
    """
    
    def run_experiment(self, prompts, openai_client=None, gemini_api_key=None):
        """
        Run inter-model consistency test
        
        Args:
            prompts: List of test prompts
            openai_client: OpenAI client object
            gemini_api_key: Gemini API key
            
        Returns:
            dict: Consistency metrics
        """
        print(f"[EXP-4] Comparing {len(prompts)} prompts across models...")
        
        responses_openai = []
        responses_gemini = []
        similarities = []
        
        for i, prompt in enumerate(prompts):
            print(f"  Prompt {i+1}/{len(prompts)}...", end=' ')
            
            # Get OpenAI response
            if openai_client:
                try:
                    resp_openai = self._get_openai_response(openai_client, prompt)
                    responses_openai.append(resp_openai)
                except Exception as e:
                    print(f"[OpenAI Error: {str(e)[:30]}]", end=' ')
                    responses_openai.append(None)
            
            # Get Gemini response
            if gemini_api_key:
                try:
                    resp_gemini = self._get_gemini_response(gemini_api_key, prompt)
                    responses_gemini.append(resp_gemini)
                except Exception as e:
                    print(f"[Gemini Error: {str(e)[:30]}]", end=' ')
                    responses_gemini.append(None)
            
            # Calculate similarity
            if responses_openai[-1] and responses_gemini[-1]:
                sim = self._calculate_similarity(responses_openai[-1], responses_gemini[-1])
                similarities.append(sim)
                print(f"Similarity={sim:.2f}")
            else:
                print("[Skipped]")
            
            time.sleep(1)  # Rate limiting
        
        avg_similarity = np.mean(similarities) if similarities else 0
        
        results = {
            'n_prompts': len(prompts),
            'similarities': similarities,
            'avg_similarity': avg_similarity,
            'openai_responses': responses_openai,
            'gemini_responses': responses_gemini,
            'timestamp': datetime.now().isoformat()
        }
        
        print(f"[EXP-4] Average Similarity: {avg_similarity:.3f}")
        return results
    
    @staticmethod
    def _get_openai_response(client, prompt, max_tokens=200):
        """Get response from OpenAI"""
        response = client.chat.completions.create(
            model="gpt-4o-mini",
            messages=[{"role": "user", "content": prompt}],
            temperature=0.3,
            max_tokens=max_tokens
        )
        return response.choices[0].message.content.strip()
    
    @staticmethod
    def _get_gemini_response(api_key, prompt, max_tokens=200):
        """Get response from Gemini"""
        import google.generativeai as genai
        genai.configure(api_key=api_key)
        model = genai.GenerativeModel('gemini-1.5-flash')
        response = model.generate_content(prompt)
        return response.text.strip()
    
    @staticmethod
    def _calculate_similarity(text1, text2):
        """Calculate text similarity (simple word overlap)"""
        words1 = set(text1.lower().split())
        words2 = set(text2.lower().split())
        
        if len(words1) == 0 or len(words2) == 0:
            return 0
        
        intersection = len(words1 & words2)
        union = len(words1 | words2)
        
        return intersection / union if union > 0 else 0
    
    def print_report(self, results):
        """Print evaluation report"""
        print("\n" + "="*70)
        print("EXPERIMENT 4: INTER-MODEL CONSISTENCY")
        print("="*70)
        
        data = [
            ["Prompts Tested", results['n_prompts']],
            ["Average Similarity", f"{results['avg_similarity']:.3f}"],
            ["Min Similarity", f"{min(results['similarities']):.3f}"],
            ["Max Similarity", f"{max(results['similarities']):.3f}"],
            ["Std Dev", f"{np.std(results['similarities']):.3f}"]
        ]
        print(tabulate(data, headers=["Metric", "Value"], tablefmt="grid"))


# =====================================================================
# EXPERIMENT 5: SCALABILITY (LOAD TESTING)
# =====================================================================

class ScalabilityEvaluation:
    """
    Simulate concurrent users and measure response times
    
    Example:
        eval = ScalabilityEvaluation()
        results = eval.simulate_concurrent_users(
            n_users=[10, 25, 50, 100],
            request_func=lambda: search_system(),
            timeout=30
        )
    """
    
    def simulate_concurrent_users(self, n_users_list, request_func, timeout=30):
        """
        Simulate concurrent user load
        
        Args:
            n_users_list: List of concurrent user counts to test
            request_func: Function that simulates one user request
            timeout: Timeout per request (seconds)
            
        Returns:
            dict: Performance metrics
        """
        print(f"[EXP-5] Simulating concurrent users: {n_users_list}")
        
        results_by_load = {}
        
        for n_users in n_users_list:
            print(f"  Testing with {n_users} concurrent users...")
            response_times = []
            errors = 0
            
            for _ in range(n_users):
                start = time.time()
                try:
                    request_func()
                    response_time = time.time() - start
                    if response_time <= timeout:
                        response_times.append(response_time)
                    else:
                        errors += 1
                except Exception as e:
                    errors += 1
            
            # Calculate statistics
            success_rate = (n_users - errors) / n_users * 100
            avg_response = np.mean(response_times) if response_times else 0
            p95_response = np.percentile(response_times, 95) if response_times else 0
            p99_response = np.percentile(response_times, 99) if response_times else 0
            
            results_by_load[n_users] = {
                'n_users': n_users,
                'success_rate': success_rate,
                'avg_response_time': avg_response,
                'p95_response_time': p95_response,
                'p99_response_time': p99_response,
                'errors': errors,
                'response_times': response_times
            }
            
            print(f"    Avg Response: {avg_response:.2f}s, "
                  f"Success Rate: {success_rate:.1f}%")
        
        return results_by_load
    
    def print_report(self, results_by_load):
        """Print scalability report"""
        print("\n" + "="*70)
        print("EXPERIMENT 5: SCALABILITY & LOAD TESTING")
        print("="*70)
        
        data = []
        for n_users in sorted(results_by_load.keys()):
            r = results_by_load[n_users]
            data.append([
                r['n_users'],
                f"{r['avg_response_time']:.2f}s",
                f"{r['p95_response_time']:.2f}s",
                f"{r['p99_response_time']:.2f}s",
                f"{r['success_rate']:.1f}%"
            ])
        
        print(tabulate(data, headers=["Users", "Avg", "P95", "P99", "Success%"], 
                      tablefmt="grid"))


# =====================================================================
# EXPERIMENT 6: USER SATISFACTION SURVEY
# =====================================================================

class UserSatisfactionEvaluation:
    """
    Collect and analyze user satisfaction survey results
    
    Example:
        eval = UserSatisfactionEvaluation()
        results = eval.run_experiment(
            survey_responses=[
                {'q1': 5, 'q2': 4, 'q3': 5, ...},
                ...
            ]
        )
    """
    
    def run_experiment(self, survey_responses):
        """
        Run user satisfaction evaluation
        
        Args:
            survey_responses: List of dicts with question responses (1-5 scale)
            
        Returns:
            dict: Satisfaction metrics
        """
        print(f"[EXP-6] Analyzing {len(survey_responses)} survey responses...")
        
        if not survey_responses:
            print("  No responses provided")
            return {}
        
        # Convert to DataFrame
        df_responses = pd.DataFrame(survey_responses)
        
        # Calculate metrics
        overall_satisfaction = df_responses.mean().mean()
        
        # NPS calculation (promoters - detractors)
        # Assuming last question is "Would you recommend?" on 0-10 scale
        nps_score = None
        if 'recommend_score' in df_responses.columns:
            recommend = df_responses['recommend_score']
            promoters = (recommend >= 9).sum() / len(recommend) * 100
            detractors = (recommend <= 6).sum() / len(recommend) * 100
            nps_score = promoters - detractors
        
        # Per-question analysis
        question_means = df_responses.mean()
        
        results = {
            'n_respondents': len(survey_responses),
            'overall_satisfaction': overall_satisfaction,
            'nps_score': nps_score,
            'question_means': question_means.to_dict(),
            'timestamp': datetime.now().isoformat()
        }
        
        print(f"[EXP-6] Overall Satisfaction: {overall_satisfaction:.2f}/5")
        if nps_score:
            print(f"[EXP-6] NPS Score: {nps_score:.1f}")
        
        return results
    
    def print_report(self, results):
        """Print user satisfaction report"""
        print("\n" + "="*70)
        print("EXPERIMENT 6: USER SATISFACTION SURVEY")
        print("="*70)
        
        data = [
            ["Respondents", results['n_respondents']],
            ["Overall Satisfaction", f"{results['overall_satisfaction']:.2f}/5.00"],
        ]
        
        if results['nps_score'] is not None:
            data.append(["NPS Score", f"{results['nps_score']:.1f}"])
        
        print(tabulate(data, headers=["Metric", "Value"], tablefmt="grid"))
        
        print("\nPer-Question Scores:")
        question_data = [[q, f"{v:.2f}/5.00"] for q, v in results['question_means'].items()]
        print(tabulate(question_data, headers=["Question", "Score"], tablefmt="grid"))


# =====================================================================
# METRICS EXPORT & REPORTING
# =====================================================================

class EvaluationReporter:
    """Export evaluation results for documentation"""
    
    @staticmethod
    def generate_summary_table(all_experiments):
        """Generate summary table of all experiments"""
        summary_data = []
        
        if 'topic_classification' in all_experiments:
            exp = all_experiments['topic_classification']
            summary_data.append([
                "Topic Classification",
                f"{exp['precision']:.1%}",
                f"{exp['recall']:.1%}",
                f"{exp['f1_score']:.3f}",
                f"{exp['n_samples']}"
            ])
        
        if 'speed_improvement' in all_experiments:
            exp = all_experiments['speed_improvement']
            summary_data.append([
                "Speed Improvement",
                f"{exp['speedup_percentage']:.1f}%",
                f"{exp['speedup_factor']:.1f}x",
                exp['system_time_formatted'],
                exp['manual_time_formatted']
            ])
        
        if 'deduplication' in all_experiments:
            exp = all_experiments['deduplication']
            summary_data.append([
                "Deduplication",
                f"{exp['precision']:.1%}",
                f"{exp['recall']:.1%}",
                f"{exp['f1_score']:.3f}",
                f"{exp['duplicates_removed']}"
            ])
        
        if 'inter_model_consistency' in all_experiments:
            exp = all_experiments['inter_model_consistency']
            summary_data.append([
                "Model Consistency",
                f"{exp['avg_similarity']:.3f}",
                "-",
                "-",
                f"{exp['n_prompts']}"
            ])
        
        if 'user_satisfaction' in all_experiments:
            exp = all_experiments['user_satisfaction']
            summary_data.append([
                "User Satisfaction",
                f"{exp['overall_satisfaction']:.2f}/5.00",
                f"NPS: {exp['nps_score']:.1f}" if exp['nps_score'] else "-",
                "-",
                f"{exp['n_respondents']} users"
            ])
        
        headers = ["Experiment", "Primary Metric", "Secondary", "Tertiary", "Sample"]
        return tabulate(summary_data, headers=headers, tablefmt="grid")
    
    @staticmethod
    def export_to_json(all_experiments, filename='evaluation_results.json'):
        """Export all results to JSON"""
        with open(filename, 'w') as f:
            json.dump(all_experiments, f, indent=2, default=str)
        print(f"Exported results to {filename}")
    
    @staticmethod
    def export_to_markdown(all_experiments, filename='evaluation_report.md'):
        """Export results to Markdown"""
        with open(filename, 'w') as f:
            f.write("# Evaluation Results\n\n")
            f.write(f"**Generated:** {datetime.now().isoformat()}\n\n")
            
            # Summary table
            f.write("## Summary\n\n")
            f.write("| Experiment | Result |\n")
            f.write("|---|---|\n")
            
            for exp_name, exp_data in all_experiments.items():
                if 'precision' in exp_data:
                    f.write(f"| {exp_name} | "
                           f"Precision: {exp_data['precision']:.1%}, "
                           f"Recall: {exp_data['recall']:.1%} |\n")
            
            f.write("\n")
            
            # Detailed results
            for exp_name, exp_data in all_experiments.items():
                f.write(f"## {exp_name}\n\n")
                f.write(f"```json\n{json.dumps(exp_data, indent=2, default=str)}\n```\n\n")
        
        print(f"Exported report to {filename}")


# =====================================================================
# QUICK START EXAMPLE
# =====================================================================

def run_all_experiments_example():
    """
    Example of running all experiments
    
    Usage:
        python evaluation_module.py
    """
    print("="*70)
    print("EVALUATION MODULE - QUICK START EXAMPLE")
    print("="*70)
    
    # Experiment 1: Topic Classification
    print("\n[1] Topic Classification Accuracy")
    eval1 = TopicClassificationEvaluation()
    # Note: Requires actual data - this is just structure
    # results1 = eval1.run_experiment(df, ground_truth_labels, n_topics=5)
    # eval1.print_report(results1)
    print("  [Example only - needs actual data]")
    
    # Experiment 2: Speed Improvement
    print("\n[2] Speed Improvement")
    eval2 = SpeedEvaluation()
    results2 = eval2.run_experiment(
        system_search_time=120,  # 2 minutes
        manual_review_time=8*24*3600  # 8 days
    )
    eval2.print_report(results2)
    
    # Experiment 3: Deduplication
    print("\n[3] Deduplication Accuracy")
    eval3 = DeduplicationEvaluation()
    # results3 = eval3.run_experiment(df)
    # eval3.print_report(results3)
    print("  [Example only - needs actual data]")
    
    # Experiment 4: Inter-Model Consistency
    print("\n[4] Inter-Model Consistency")
    eval4 = InterModelConsistencyEvaluation()
    print("  [Requires OpenAI & Gemini API keys]")
    
    # Experiment 5: Scalability
    print("\n[5] Scalability Testing")
    eval5 = ScalabilityEvaluation()
    # Example mock request
    mock_results = eval5.simulate_concurrent_users(
        n_users_list=[10, 25, 50],
        request_func=lambda: time.sleep(np.random.uniform(1, 3)),  # Mock processing
        timeout=30
    )
    eval5.print_report(mock_results)
    
    # Experiment 6: User Satisfaction
    print("\n[6] User Satisfaction Survey")
    eval6 = UserSatisfactionEvaluation()
    # Mock survey responses
    mock_survey = [
        {'q1': 5, 'q2': 4, 'q3': 5, 'q4': 4, 'q5': 5, 'recommend_score': 9},
        {'q1': 4, 'q2': 5, 'q3': 4, 'q4': 5, 'q5': 4, 'recommend_score': 8},
    ]
    results6 = eval6.run_experiment(mock_survey)
    eval6.print_report(results6)
    
    print("\n" + "="*70)
    print("Evaluation module ready to use!")
    print("="*70)


if __name__ == "__main__":
    run_all_experiments_example()