Hack Session: Evaluating ML Models for Bias – Build an Interpretable Model using a Financial Dataset

Nov 14, 2019

11:30

Auditorium 2

60 minutes

Machine Learning

Machine learning models are increasingly used to inform high stakes decisions about people. Although machine learning, by its very nature, is always a form of statistical discrimination, the discrimination becomes objectionable when it places certain privileged groups at a systematic advantage and certain unprivileged groups at a systematic disadvantage.

In this hack session, we will discuss the concepts and capabilities of a model to test for biases and explanations. The wider spectrum of explainability methods, notably data explanations, metrics and persona specific explanations will be briefed.

Structure of the hack session

  • Implement an ML model using a Financial dataset and fit it with known algorithms like Random Forest and evaluate the Model
  • Persist the Model for future reference
  • Setup a datamart and bind it with the ML engines
  • Score the Model so that the monitors are configured
  • Enable quality monitoring, feedback logging, and fairness monitoring
  • Through historical performance metrics identify the transactions for Explainability
  • Have a visual and text view of bias and model explanation
  • Open discussion to implement the above flow for another dataset

Key Takeaways from the Hack Session

  • Capability to implement a similar solution for a different dataset
  • Reference notebook to be shared
  • A new dataset for exploration and the implementation guide will be given

 

Check out the video below to know more about the session.

  • Prateek Goyal

    Software Engineer

    IBM Watson OpenScale

  • Rajesh K Jeyapaul

    Senior AI Architect

    IBM

Copyright 2019 Analytics Vidhya. All rights reserved