MIT EECS — Lincoln Labs Undergraduate Research and Innovation Scholar
Big Data and Missing Data in Omnichannel Pricing
Our research considers how different channels (such as brick-and-mortar and online) as well as different sources of data (such as transaction and social media data) can help improve demand prediction and pricing decisions. One challenge in predicting demand is the problem of missing data: the retailer observes the amount of rewards offered to customers who make purchases, but does not observe any data on customers who decide not to buy. The Expectation-Maximization (EM) algorithm is a classical approach to parameter estimation with missing data, but assumes a certain shape of the distribution of rewards. As a non-parametric alternative, we have proposed the Non-Parametric Maximization (“NPM”) algorithm. The project will study the NPM algorithm both analytically and through simulations.
I have worked as a UROP student with Professor Perakis for one and a half years. My SuperUROP project is a direct extension to one of my projects in the past. I am interested in Operations Research as it deals with the application of mathematics and computer science to solve practical decision-making problems.