Ishan  Pakuwal

Ishan Pakuwal

Scholar Title

MIT EECS Undergraduate Research and Innovation Scholar

Research Title

Bias and Robustness of ML Models Against Adversarial Attacks in Finance

Cohort

2021–2022

Department

EECS

Research Areas
  • Artificial Intelligence & Machine Learning
Supervisor

Luca Daniel

Abstract

Adoption of Machine Learning models in different applications in academia, medicine and even fields such
as finance has skyrocketed over the years. As the adoption of these models is quickly entering into critical
applications, attention needs to be paid to ensure that the models’ decisions can be explained and that
they do not discriminate solely based on factors such as gender, religion and others. An important critical
domain of this application is in finance particularly in loan classification where we do not want models
to predict default based on biases. Since most ML models essentially function as black box models, this
increases the chance that some of the models that get employed by corporations may be biased. Therefore
to hold corporations accountable if biased decisions are made, it is necessary to explore the explainability
of ML model in explainability and this is what our paper explores explainability through methods such as
logistic rule regression model and CatBoost Model.

Back to Scholars