|
import pandas as pd
|
|
from sklearn.ensemble import IsolationForest
|
|
from smolagents import CodeAgent
|
|
|
|
|
|
class FraudDetectionAgent:
|
|
pass
|
|
|
|
|
|
CodeAgent = FraudDetectionAgent()
|
|
|
|
|
|
class TransactionModel:
|
|
def __init__(self, transaction_id: int, amount: float, timestamp: str, location_lat: float, location_long: float):
|
|
self.transaction_id = transaction_id
|
|
self.amount = amount
|
|
self.timestamp = timestamp
|
|
self.location_lat = location_lat
|
|
self.location_long = location_long
|
|
|
|
def to_dict(self):
|
|
return {
|
|
"transaction_id": self.transaction_id,
|
|
"amount": self.amount,
|
|
"timestamp": self.timestamp,
|
|
"location_lat": self.location_lat,
|
|
"location_long": self.location_long
|
|
}
|
|
|
|
|
|
class FraudResult:
|
|
def __init__(self, transaction_id: int, amount: float, timestamp: str, anomaly_score: int):
|
|
self.transaction_id = transaction_id
|
|
self.amount = amount
|
|
self.timestamp = timestamp
|
|
self.anomaly_score = anomaly_score
|
|
|
|
def to_dict(self):
|
|
return {
|
|
"transaction_id": self.transaction_id,
|
|
"amount": self.amount,
|
|
"timestamp": self.timestamp,
|
|
"anomaly_score": self.anomaly_score
|
|
}
|
|
|
|
|
|
class FraudDetectionAgent(FraudDetectionAgent):
|
|
def __init__(self, data_path: str):
|
|
super().__init__()
|
|
self.data_path = data_path
|
|
self.df = None
|
|
self.X = None
|
|
self.model = IsolationForest(contamination=0.01, random_state=42)
|
|
|
|
def load_data(self):
|
|
self.df = pd.read_csv(self.data_path)
|
|
print(f"Loaded {len(self.df)} transactions.")
|
|
|
|
def preprocess(self):
|
|
self.df['transaction_hour'] = pd.to_datetime(self.df['timestamp']).dt.hour
|
|
features = ['amount', 'transaction_hour', 'location_lat', 'location_long']
|
|
self.df = self.df.dropna(subset=features)
|
|
self.X = self.df[features]
|
|
|
|
def detect_fraud(self):
|
|
self.df['anomaly_score'] = self.model.fit_predict(self.X)
|
|
frauds = self.df[self.df['anomaly_score'] == -1]
|
|
print(f"Detected {len(frauds)} potential fraudulent transactions.")
|
|
return [
|
|
FraudResult(
|
|
row['transaction_id'],
|
|
row['amount'],
|
|
row['timestamp'],
|
|
row['anomaly_score']
|
|
).to_dict()
|
|
for _, row in frauds.iterrows()
|
|
]
|
|
|
|
def run(self):
|
|
self.load_data()
|
|
self.preprocess()
|
|
return self.detect_fraud()
|
|
|
|
|
|
if __name__ == "__main__":
|
|
agent = FraudDetectionAgent(data_path="transactions.csv")
|
|
agent.run()
|
|
print("\nFraud detection completed.")
|
|
|
|
|