Fraud detection algorithms in action

Fraud Detection in Python

Charlotte Werger

Data Scientist

Traditional fraud detection with rules based systems

Drawbacks of using rules based systems

Rules based systems have their limitations:

Fixed thresholds per rule to determine fraud
Limited to yes/no outcomes
Fail to capture interaction between features

Why use machine learning for fraud detection?

Machine learning models adapt to the data, and thus can change over time
Uses all the data combined rather than a threshold per feature
Can give a score, rather than a yes/no
Will typically have a better performance and can be combined with rules

Refresher on machine learning models

from sklearn.linear_model import LinearRegression
from sklearn.model_selection import train_test_split
from sklearn import metrics

# Step 1: split your features and labels into train and test data 
X_train, X_test, y_train, y_test = train_test_split(X, y, test_size=0.2)

# Step 2: Define which model you want to use 
model = LinearRegression()

# Step 3: Fit the model to your training data 
model.fit(X_train, y_train)

# Step 4: Obtain model predictions from your test data
y_predicted = model.predict(X_test)

# Step 5: Compare y_test to predictions and obtain performance metrics
print (metrics.r2_score(y_test, y_predicted))

0.821206237313

What you'll be doing in the upcoming chapters

Chapter 2. Supervised learning: train a model using existing fraud labels
Chapter 3. Unsupervised learning: use your data to determine what is 'suspicious' behavior without labels
Chapter 4. Fraud detection using text data: Learn how to augment your fraud detection models with text mining and topic modeling

Let's practice!

Fraud Detection in Python