vix.ing · top · new · best · stats · spec

Discovering Interpretable Machine Learning Models in Parallel\n Coordinates

2021/06/14 by Boris Kovalerchuk, Kovalerchuk, Boris, Dustin Hayes +1 · 1 citation
Computer Science · #Data Visualization and Analytics #FOS: Computer and information sciences #Machine Learning (cs.LG) #Neural Networks and Applications

paper · pdf · doi:10.48550/arxiv.2106.07474

openalex publication_date 2021/06/14 · openalex created_date 2025/10/10 · openalex updated_date 2026/07/28

Abstract

This paper contributes to interpretable machine learning via visual knowledge\ndiscovery in parallel coordinates. The concepts of hypercubes and hyper-blocks\nare used as easily understandable by end-users in the visual form in parallel\ncoordinates. The Hyper algorithm for classification with mixed and pure\nhyper-blocks (HBs) is proposed to discover hyper-blocks interactively and\nautomatically in individual, multiple, overlapping, and non-overlapping\nsetting. The combination of hyper-blocks with linguistic description of visual\npatterns is presented too. It is shown that Hyper models generalize decision\ntrees. The Hyper algorithm was tested on the benchmark data from UCI ML\nrepository. It allowed discovering pure and mixed HBs with all data and then\nwith 10-fold cross validation. The links between hyper-blocks, dimension\nreduction and visualization are established. Major benefits of hyper-block\ntechnology and the Hyper algorithm are in their ability to discover and observe\nhyper-blocks by end-users including side by side visualizations making patterns\nvisible for all classes. Another advantage of sets of HBs relative to the\ndecision trees is the ability to avoid both data overgeneralization and\noverfitting.\n

Cited by

Related