Location via proxy:   [ UP ]  
[Report a bug]   [Manage cookies]                
SlideShare a Scribd company logo
1
Arnaud Wald
MACHINE LEARNING ENGINEER
FRAUD DETECTION
WITH MACHINE LEARNING
2
WHAT IS FRAUD ?
3
4
Stolen Credit Card → Unpaid resources
FRAUD @ SCALEWAY
DEALING WITH FRAUD
Quotas Manual
Intervention
5
6
Frequency Magnitude Sophistication
INCREASING TRENDS
THE IDEAL SOLUTION
7
While having a low false positive rate
Fast Scalable Adaptable
IS THERE A SOLUTION?
8
MACHINE LEARNING
9
While having a low false posiJve rate
Fast Scalable Adaptable✓ ✓ ✓
✓
MACH IN E LEARN IN G
PRO JECT
10
OBJECTIVE
11
ML
MODEL
Fraud
Non Fraud
New User ACTION
MACHINE LEARNING PIPELINE
12
ML
MODEL
Fraud
Non Fraud
New User ACTION
ML PIPELINE
13
Data
Preparation
Model Building
& Training
Model
Deployment
WHERE DO WE GET THE DATA ?
14
GATHERING DATA
15
Email
ToS
Phone
Address
Credit
Card
2FA
Start
Instance
SSH
CHOOSING A
MACHINE LEARNING MODEL
16
Deep Learning
- Very Powerful
- Needs lots (millions) of data
Classic ML:
Random Forest
Gradient Boosting
Choosing a Machine Learning Algorithm
17
MODEL SELECTION
Chosen algorithm: CatBoost
- Number of trees
- Depth of trees
- Learning rate
18
HYPERPARAMETERS
Example from Titanic data
Source: Wikipedia
CatBoost brings:
- Powerful
- Open-source
- Handles categorical data very well
- Automatic overfitting detection
- (Works on GPU)
19
CATBOOST
EVALUATING PERFORMANCE
20
Trade-off between being too severe/ Being too friendly
Avoiding Overfitting
False positives > Innocents wrongly flagged
False Negatives > Fraudsters not detected
EVALUATING PERFORMANCE
Hyperparameter Tuning
21
MACHINE LEARNING PIPELINE
22
ML
MODEL
Fraud
Non Fraud
New User ACTION
GOING FURTHER
23
NEW PRO D U CTS
24
OTHER FRAUD BEHAVIORS
Stay tuned for exclusive tutorials and updates,
follow us on Twitter and LinkedIn @Scaleway
THANK YOU !
And follow me on LinkedIn
@ArnaudWald
26

More Related Content

Fraud detection with Machine Learning