A typical organization loses an estimated 5% of its yearly revenue to fraud. In this course, you will learn how to fight fraud by using data. For example, you'll learn how to apply supervised learning algorithms to detect fraudulent behavior similar to past ones, as well as unsupervised learning methods to discover new types of fraud activities. Moreover, in fraud analytics you often deal with highly imbalanced datasets when classifying fraud versus non-fraud, and during this course you will pick up some techniques on how to deal with that. The course provides a mix of technical and theoretical insights and shows you hands-on how to practically implement fraud detection models. In addition, you will get tips and advice from real-life experience to help you prevent making common mistakes in fraud analytics.
Introduction and preparing your dataFree
In this chapter, you'll learn about the typical challenges associated with fraud detection, and will learn how to resample your data in a smart way, to tackle problems with imbalanced data.Introduction to fraud detection50 xpChecking the fraud to non-fraud ratio100 xpPlotting your data100 xpIncreasing successful detections using data resampling50 xpResampling methods for imbalanced data50 xpApplying SMOTE100 xpCompare SMOTE to original data100 xpFraud detection algorithms in action50 xpExploring the traditional way to catch fraud100 xpUsing ML classification to catch fraud100 xpLogistic regression combined with SMOTE100 xpUsing a pipeline100 xp
Fraud detection using labeled data
Now that you're familiar with the main challenges of fraud detection, you're about to learn how to flag fraudulent transactions with supervised learning. You will use classifiers, adjust them, and compare them to find the most efficient fraud detection model.Review of classification methods50 xpNatural hit rate100 xpRandom Forest Classifier - part 1100 xpRandom Forest Classifier - part 2100 xpPerformance evaluation50 xpPerformance metrics for the RF model100 xpPlotting the Precision Recall Curve100 xpAdjusting your algorithm weights50 xpModel adjustments100 xpAdjusting your Random Forest to fraud detection100 xpGridSearchCV to find optimal parameters100 xpModel results using GridSearchCV100 xpEnsemble methods50 xpLogistic Regression100 xpVoting Classifier100 xpAdjust weights within the Voting Classifier100 xp
Fraud detection using unlabeled data
This chapter focuses on using unsupervised learning techniques to detect fraud. You will segment customers, use K-means clustering and other clustering algorithms to find suspicious occurrences in your data.Normal versus abnormal behavior50 xpExploring your data100 xpCustomer segmentation100 xpUsing statistics to define normal behavior100 xpClustering methods to detect fraud50 xpScaling the data100 xpK-means clustering100 xpElbow method100 xpAssigning fraud versus non-fraud50 xpDetecting outliers100 xpChecking model results100 xpOther clustering fraud detection methods50 xpDBSCAN100 xpAssessing smallest clusters100 xpChecking results100 xp
Fraud detection using text
In this final chapter, you will use text data, text mining, and topic modeling to detect fraudulent behavior.Using text data50 xpWord search with dataframes100 xpUsing list of terms100 xpCreating a flag100 xpText mining to detect fraud50 xpRemoving stopwords100 xpCleaning text data100 xpTopic modeling on fraud50 xpCreate dictionary and corpus100 xpLDA model100 xpFlagging fraud based on topics50 xpInterpreting the topic model50 xpFinding fraudsters based on topic100 xpRecap50 xp
Charlotte WergerSee More
Director of Advanced Analytics at Nike
Dr. Charlotte Werger currently works at Nike as a Director of Advanced Analytics. Charlotte is a data scientist with a background in econometrics and finance. She loves applying Machine Learning to a broad variety of problems, ranging from image recognition to fraud detection, to customer recommender systems. Charlotte has previously worked in finance as Head of Data Science at Van Lanschot Kempen, and as a quantitative researcher and portfolio manager for BlackRock and Man AHL. In those roles, she specialized in using data science to predict movements in stock markets. As the former Head of Education at Faculty, she loves teaching data science on- and off-line. Charlotte is also active as a Data Science mentor for the Springboard program. Charlotte holds a P.h.D from the European University Institute.