University of Twente Student Theses

Login

Visualizing feature importance dependencies in machine learning

Koehorst, G.S. (2021) Visualizing feature importance dependencies in machine learning.

[img] PDF
1MB
Abstract:There is a growing need for methods that can help to ex-plain machine learning models. Feature importance analysis is often used to determine the relationship between a feature and the target of the model. However, this analysis does not take into account the dependencies between features. This research defines three model-agnostic methods to visualize feature importance dependencies. The defined methods are evaluated using synthetic data. The proposed methods are a feature importance dependency bar chart, holding out on correlation pie-charts, and a feature importance dependency tree. All three methods were able to effectively visualize feature importance dependencies when the number of features was six or lower. When the number of features was eight, the tree method became less effective. However, we conclude that all three methods were able to visualize feature importance dependencies. So, the contribution of this paper is that we define and evaluate three approaches to visualize feature importance dependencies, which can help us in showing how dependencies affect feature importances, in order to improve the explainability of machine learning.
Item Type:Essay (Bachelor)
Faculty:EEMCS: Electrical Engineering, Mathematics and Computer Science
Subject:54 computer science
Programme:Business & IT BSc (56066)
Link to this item:http://purl.utwente.nl/essays/86897
Export this item as:BibTeX
EndNote
HTML Citation
Reference Manager

 

Repository Staff Only: item control page