Event details from: Business Analytics Seminar: Behrouz Tangabi (ING)

Date
3 October 2022
Time
12:00 – 13:00

Introduction

More data usually provides more opportunities and fewer challenges to learn, while with less data, the story is other way around. In classification learning problems, data containing more information from one class than others is known as imbalanced data, and therefore, relevant learning is known as imbalanced classification. The main challenge in imbalanced learning is that ML models tend to be biased towards the majority class and misclassify the minority class. Hence, building ML models with optimal performance on imbalanced datasets is not an easy task and requires a careful strategy. Cost-sensitive algorithms try to solve the imbalanced classification by assigning different weights to each class. Some models, such as logistic regression, are sensitive to class weight changes, while some others not. A nonlinear (shock-like) relationship between the class weights and the model performance is seen in tree-based models. The purpose of this talk is to discuss the effect of class-weight change in tree-based algorithms with a more focus on splitting decision metrics. The talk also seeks to briefly address other aspects in imbalanced classification.

General information

This seminar will be organised in a hybrid setup. If you are interested in joining this seminar, please send an email to the secretariat of Amsterdam Business School at secbs-abs@uva.nl.