Summary  
This chapter demonstrates how to use an IsolationForest model from scikit-learn to detect anomalies by selecting numerical features from a pandas DataFrame, fitting the model, predicting outliers, and printing flagged anomalies.

General domain of usage  
Compliance monitoring

Machine learning is a valuable tool in compliance because it helps you detect patterns and predict suspicious activity more efficiently than manual review alone. In a compliance context, machine learning can analyze large volumes of transaction data to identify behaviors that deviate from the norm, which may indicate fraud, money laundering, or other regulatory breaches. By learning from historical data, these algorithms can flag unusual transactions for further investigation, reducing the risk of missed violations and improving the speed and accuracy of compliance monitoring.

from sklearn.ensemble import IsolationForest
import pandas as pd

# Hardcoded dataset of transactions
data = {
    "transaction_id": [1, 2, 3, 4, 5, 6, 7, 8],
    "amount": [100, 110, 95, 102, 5000, 98, 105, 120],
    "customer_age": [34, 36, 33, 35, 40, 34, 37, 32]
}

df = pd.DataFrame(data)

# Use only numerical features for anomaly detection
features = df[["amount", "customer_age"]]

# Create and fit IsolationForest model
model = IsolationForest(contamination=0.125, random_state=42)
df["anomaly"] = model.fit_predict(features)

The **IsolationForest** algorithm works by randomly selecting features and splitting data points to isolate observations. In practice, **outliers**—or anomalies—are easier to isolate because they differ significantly from the rest of the data. This makes `IsolationForest` well-suited for compliance data, where you want to flag rare or unusual transactions that could signal suspicious activity. By training on features such as `transaction amount` and `customer_age`, the model can help you focus your review on the most atypical transactions, improving the efficiency and effectiveness of compliance monitoring.

# Print out transactions flagged as outliers
outliers = df[df["anomaly"] == -1]
print("Flagged outlier transactions:")
print(outliers[["transaction_id", "amount", "customer_age"]])

What is the purpose of using machine learning in compliance?

Which scikit-learn model can be used for anomaly detection?

What is an outlier in the context of compliance?

Focused on advanced automation and monitoring of compliance in MLOps using Python. Covers rule-based compliance systems, anomaly detection, and regulatory validation most relevant to security and governance in machine learning operations.

Simple Machine Learning for Compliance

1. What is the purpose of using machine learning in compliance?

2. Which scikit-learn model can be used for anomaly detection?

3. What is an outlier in the context of compliance?