SMUBIA x DSSG Part 1
We're excited to collaborate with SMUBIA to invite Prof Richard Crowley from SMU and Akanksha from Rakuten to give our second talk for 2019!
Speaker:
Richard Crowley joined Singapore Management University in 2016. He received his PhD in Accountancy from the University of Illinois Urbana-Champaign and received Bachelor's degrees in Accountancy, Finance, and Theoretical Mathematics from the University of Illinois Urbana-Champaign in 2012. His research examines financial accounting using both archival and analytical methods. Much of his archival work deals with large sets of unstructured data (e.g., textual disclosures) using high-powered computing algorithms to address accounting issues that are otherwise infeasible to approach.
https://www.smu.edu.sg/faculty/profile/144766/Richard-CROWLEY
https://rmc.link/author/dr-richard-m-crowley.html
Abstract:
Large corporate frauds can individually wipe out billions of dollars of investment, harming financial stability and destroying confidence in our financial systems. In this talk, I will discuss my research on using topic modeling to detect financial misreporting, including our experimental validation of Latent Dirichlet Allocation and the econometrics behind our model. Our proposed model extends traditional fraud detection techniques by analyzing what companies discuss in the text of their annual reports. Building on theory from psychology and communications, we expect and find that companies appear to intentionally change what they report when committing fraud. Our research finds that, for the biggest frauds, our proposed model can improve detection rates by 59% over the best traditional model of financial misreporting. The psychological nature of the approach, combined with an ever-shifting benchmark, makes our detection algorithm more robust to gaming than traditional models. This talk is based on research available publicly at https://ssrn.com/abstract=2803733. Slides will be available at https://rmc.link/DSSG as of the time of the talk.
Speaker:
Akanksha Tiwari works at Rakuten Institute of Technology (RIT), Singapore, a dedicated R&D organization for the Rakuten Group, on problems specific to user behavior modeling. She has worked closely with Rakuten Viki's (Rakuten’s video streaming platform for Korean and Asian content) growth and monetization teams to drive data exploration, tests, and models; present data-driven insights and actionable recommendations. Part of her work is to work with product and marketing to reduce churn and drive subscriptions using subscription churn and upsell prediction models.
https://www.linkedin.com/in/akanksha-tiwari-3737666a
Abstract:
In this talk, Akanksha will focus specifically on how the data team has been leveraging machine interpretation techniques such as SHAP values, to develop a deeper understanding of key drivers of user behavior.
Help us caption & translate this video!
