5293 FINAL PROJECT

1 Introduction

In this project, we investigate the robustness of model explanations across three dimensions: randomness in model training, size of training data, and consistency between model-agnostic explanations and inherently interpretable models. We use Random Forest and Logistic Regression to test these three aspects, and the process is structured in a 12 steps:

[Step1 Chepter]

Repeatability - Random Forest

Repeatability Analysis - Random Forest

Repeatability - Logistic Regression

Repeatability Analysis - Logistic Regression

[Step2 Chepter]

Change Training Dataset Size - Random Forest

Change Training Dataset Size Analysis - Random Forest

Change Training Dataset Size - Logistic Regression

Change Training Dataset Size Analysis - Logistic Regression

[Step3 Chepter]

Consistency - Random Forest

Consistency Analysis - Random Forest

Consistency - Logistic Regression

Consistency Analysis - Logistic Regression