Spaces:
Sleeping
Sleeping
File size: 10,541 Bytes
fd99b61 |
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 99 100 101 102 103 104 105 106 107 108 109 110 111 112 113 114 115 116 117 118 119 120 121 122 123 124 125 126 127 128 129 130 131 132 133 134 135 136 137 138 139 140 141 142 143 144 145 146 147 148 149 150 151 152 153 154 155 156 157 158 159 160 161 162 163 164 165 166 167 168 169 170 171 172 173 174 175 176 177 178 179 180 181 182 183 184 185 186 187 188 189 190 191 192 193 194 195 196 197 198 199 200 201 202 203 204 205 206 207 208 209 210 211 212 213 214 215 216 217 218 219 220 221 222 223 224 225 226 227 228 229 230 231 232 233 234 235 236 237 238 239 240 241 242 243 244 245 246 247 248 249 250 251 252 253 254 255 256 257 258 259 260 261 262 263 264 265 266 267 268 269 270 271 272 273 274 275 276 277 278 279 |
"""CSV document generator for RAG system."""
import logging
from pathlib import Path
from typing import List, Dict, Any
import pandas as pd
from langchain_core.documents import Document
logger = logging.getLogger(__name__)
class CSVDocumentGenerator:
"""Generate documents from CSV data for RAG system."""
def __init__(self, csv_path: Path, sample_size: int = 1050000) -> None:
"""Initialize CSV document generator.
Args:
csv_path: Path to the CSV file.
sample_size: Number of rows to sample from CSV (to handle large files).
"""
self.csv_path = Path(csv_path)
self.sample_size = sample_size
self.df: pd.DataFrame = None
def load_data(self) -> None:
"""Load CSV data with sampling for efficiency."""
if not self.csv_path.exists():
raise FileNotFoundError(f"CSV file not found: {self.csv_path}")
try:
logger.info(f"Loading CSV data from {self.csv_path}")
# Load with sampling to handle large file
self.df = pd.read_csv(self.csv_path, nrows=self.sample_size)
# Clean merchant names (remove 'fraud_' prefix common in synthetic datasets)
if 'merchant' in self.df.columns:
self.df['merchant'] = self.df['merchant'].str.replace('fraud_', '', regex=False)
logger.info(f"Loaded {len(self.df)} rows from CSV (merchant names cleaned)")
except Exception as e:
logger.error(f"Error loading CSV: {str(e)}")
raise
def generate_fraud_pattern_documents(self) -> List[Document]:
"""Generate documents about fraud patterns by category.
Returns:
List of documents containing fraud pattern insights.
"""
if self.df is None:
self.load_data()
documents = []
# Fraud patterns by category
category_fraud = self.df.groupby('category').agg({
'is_fraud': ['sum', 'mean', 'count']
}).round(4)
for category in category_fraud.index:
fraud_count = int(category_fraud.loc[category, ('is_fraud', 'sum')])
fraud_rate = float(category_fraud.loc[category, ('is_fraud', 'mean')] * 100)
total_txns = int(category_fraud.loc[category, ('is_fraud', 'count')])
content = f"""Fraud Pattern Analysis - Category: {category}
Based on historical transaction data analysis:
- Total Transactions: {total_txns:,}
- Fraud Cases: {fraud_count:,}
- Fraud Rate: {fraud_rate:.2f}%
- Risk Level: {'HIGH' if fraud_rate > 5 else 'MEDIUM' if fraud_rate > 1 else 'LOW'}
This category shows {'significant' if fraud_rate > 5 else 'moderate' if fraud_rate > 1 else 'low'} fraud activity in the historical dataset.
"""
documents.append(Document(
page_content=content,
metadata={
"source": "fraudTrain.csv",
"type": "fraud_pattern",
"category": category,
"fraud_rate": fraud_rate
}
))
logger.info(f"Generated {len(documents)} category fraud pattern documents")
return documents
def generate_statistical_summaries(self) -> List[Document]:
"""Generate statistical summary documents.
Returns:
List of documents containing statistical insights.
"""
if self.df is None:
self.load_data()
documents = []
# Overall statistics
total_txns = len(self.df)
fraud_txns = int(self.df['is_fraud'].sum())
fraud_rate = float(self.df['is_fraud'].mean() * 100)
avg_amount = float(self.df['amt'].mean())
fraud_avg_amount = float(self.df[self.df['is_fraud'] == 1]['amt'].mean())
legit_avg_amount = float(self.df[self.df['is_fraud'] == 0]['amt'].mean())
overall_summary = f"""Overall Fraud Detection Statistics
Dataset Summary:
- Total Transactions Analyzed: {total_txns:,}
- Fraudulent Transactions: {fraud_txns:,}
- Overall Fraud Rate: {fraud_rate:.2f}%
- Average Transaction Amount: ${avg_amount:.2f}
- Average Fraud Amount: ${fraud_avg_amount:.2f}
- Average Legitimate Amount: ${legit_avg_amount:.2f}
Key Insight: Fraudulent transactions have an average amount of ${fraud_avg_amount:.2f} compared to ${legit_avg_amount:.2f} for legitimate transactions.
"""
documents.append(Document(
page_content=overall_summary,
metadata={
"source": "fraudTrain.csv",
"type": "statistical_summary",
"scope": "overall"
}
))
# Amount range analysis
amount_bins = [0, 10, 50, 100, 500, 1000, float('inf')]
amount_labels = ['$0-10', '$10-50', '$50-100', '$100-500', '$500-1000', '$1000+']
self.df['amount_range'] = pd.cut(self.df['amt'], bins=amount_bins, labels=amount_labels)
amount_fraud = self.df.groupby('amount_range', observed=True).agg({
'is_fraud': ['sum', 'mean', 'count']
}).round(4)
amount_content = "Fraud Patterns by Transaction Amount\n\n"
for amt_range in amount_labels:
if amt_range in amount_fraud.index:
fraud_count = int(amount_fraud.loc[amt_range, ('is_fraud', 'sum')])
fraud_rate = float(amount_fraud.loc[amt_range, ('is_fraud', 'mean')] * 100)
total = int(amount_fraud.loc[amt_range, ('is_fraud', 'count')])
amount_content += f"""
Amount Range: {amt_range}
- Total Transactions: {total:,}
- Fraud Cases: {fraud_count:,}
- Fraud Rate: {fraud_rate:.2f}%
"""
documents.append(Document(
page_content=amount_content,
metadata={
"source": "fraudTrain.csv",
"type": "statistical_summary",
"scope": "amount_analysis"
}
))
logger.info(f"Generated {len(documents)} statistical summary documents")
return documents
def generate_merchant_profiles(self) -> List[Document]:
"""Generate merchant risk profile documents.
Returns:
List of documents containing merchant insights.
"""
if self.df is None:
self.load_data()
documents = []
# Top merchants by transaction volume
merchant_stats = self.df.groupby('merchant').agg({
'is_fraud': ['sum', 'mean', 'count'],
'amt': 'mean'
}).round(4)
# Get top 20 merchants by volume
top_merchants = merchant_stats.nlargest(20, ('is_fraud', 'count'))
for merchant in top_merchants.index:
fraud_count = int(top_merchants.loc[merchant, ('is_fraud', 'sum')])
fraud_rate = float(top_merchants.loc[merchant, ('is_fraud', 'mean')] * 100)
total_txns = int(top_merchants.loc[merchant, ('is_fraud', 'count')])
avg_amt = float(top_merchants.loc[merchant, ('amt', 'mean')])
content = f"""Merchant Risk Profile: {merchant}
Transaction Analysis:
- Total Transactions: {total_txns:,}
- Fraudulent Transactions: {fraud_count:,}
- Fraud Rate: {fraud_rate:.2f}%
- Average Transaction Amount: ${avg_amt:.2f}
- Risk Assessment: {'HIGH RISK' if fraud_rate > 10 else 'MEDIUM RISK' if fraud_rate > 5 else 'LOW RISK'}
This merchant profile is based on historical transaction patterns and can help identify similar fraud patterns.
"""
documents.append(Document(
page_content=content,
metadata={
"source": "fraudTrain.csv",
"type": "merchant_profile",
"merchant": merchant,
"fraud_rate": fraud_rate
}
))
logger.info(f"Generated {len(documents)} merchant profile documents")
return documents
def generate_location_insights(self) -> List[Document]:
"""Generate location-based fraud insights.
Returns:
List of documents containing location insights.
"""
if self.df is None:
self.load_data()
documents = []
# State-level analysis
state_fraud = self.df.groupby('state').agg({
'is_fraud': ['sum', 'mean', 'count']
}).round(4)
# Get top 15 states by transaction volume
top_states = state_fraud.nlargest(15, ('is_fraud', 'count'))
for state in top_states.index:
fraud_count = int(top_states.loc[state, ('is_fraud', 'sum')])
fraud_rate = float(top_states.loc[state, ('is_fraud', 'mean')] * 100)
total_txns = int(top_states.loc[state, ('is_fraud', 'count')])
content = f"""Geographic Fraud Analysis - State: {state}
Location-based Fraud Patterns:
- Total Transactions: {total_txns:,}
- Fraud Cases: {fraud_count:,}
- Fraud Rate: {fraud_rate:.2f}%
- Geographic Risk Level: {'HIGH' if fraud_rate > 5 else 'MEDIUM' if fraud_rate > 2 else 'LOW'}
This geographic area shows {'elevated' if fraud_rate > 5 else 'moderate' if fraud_rate > 2 else 'normal'} fraud activity levels.
"""
documents.append(Document(
page_content=content,
metadata={
"source": "fraudTrain.csv",
"type": "location_insight",
"state": state,
"fraud_rate": fraud_rate
}
))
logger.info(f"Generated {len(documents)} location insight documents")
return documents
def generate_all_documents(self) -> List[Document]:
"""Generate all types of documents from CSV data.
Returns:
List of all generated documents.
"""
all_documents = []
logger.info("Generating all document types from CSV data...")
all_documents.extend(self.generate_fraud_pattern_documents())
all_documents.extend(self.generate_statistical_summaries())
all_documents.extend(self.generate_merchant_profiles())
all_documents.extend(self.generate_location_insights())
logger.info(f"Generated total of {len(all_documents)} documents from CSV data")
return all_documents
|