Speaker "Nitesh Kumar" Details Back



Fraud or not Fraud? That is the question. Using modern Machine Learning techniques to identify fraud in credit card transactions.


Fraud or not Fraud? That is the question. In the talk, we will outline modern techniques on how to identify additional fraudulent transactions on Big Data and compare this to traditional fraud detection through a fictional credit card company, FraudX. In this talk, we will examine credit card fraud and how machine learning on Big Data can flag future fraudulent transactions. We will show how to select the “best models” based on multiple metrics like the associated cost of fraud, area under the ROC curve etc. We will compare machine learning based fraud detection to traditional fraud detection methods, translating the benefit into cost savings. We will also outline state of the art techniques that can be used to automatically obtain the best model given the data and the objective function without going through a series of tuning runs, truly democratizing data science and machine learning.


Nitesh is the Director of Data Science at Affirm where he is responsible for the online decisioning system that provides loans to help consumers make online purchases. Before Affirm, Nitesh was building machine learning algorithms to help democratize data science at Skytree. He holds a PhD in mathematical finance and has been practicing analytics and data science for over 9 years. In his free time, he dabbles in the stock market and bets on rare events.