Approaches for Multi-View Redescription Mining

Matej Mihelcic,Tomislav Smuc

doi:10.1109/access.2021.3054245

Matej Mihelcic, Tomislav Smuc

Open Access

https://doi.org/10.1109/access.2021.3054245

Copy DOI

Abstract

The task of redescription mining explores ways to re-describe different subsets of entities contained in a dataset and to reveal non-trivial associations between different subsets of attributes, called views. This interesting and challenging task is encountered in different scientific fields, and is addressed by a number of approaches that obtain redescriptions and allow for the exploration and analyses of attribute associations. The main limitation of existing approaches to this task is their inability to use more than two views. Our work alleviates this drawback. We present a memory efficient, extensible multi-view redescription mining framework that can be used to relate multiple, i.e. more than two views, disjoint sets of attributes describing one set of entities. The framework can use any multi-target regression or multi-label classification algorithm, with models that can be represented as sets of rules, to generate redescriptions. Multi-view redescriptions are built using incremental view-extending heuristic from initially created two-view redescriptions. In this work, we use different types of Predictive Clustering trees algorithms (regular, extra, with random output selection) and the Random Forest thereof in order to improve the quality of final redescription sets and/or execution time needed to generate them. We provide multiple performance analyses of the proposed framework and compare it against the naive approach to multi-view redescription mining. We demonstrate the usefulness of the proposed multi-view extension on several datasets, including a use-case on understanding of machine learning models - a topic of growing importance in machine learning and artificial intelligence in general.

Highlights

Redescription mining [30] aims to find multiple characterizations of different subsets of entities in a set of available data, i.e. to identify subsets of instances that can be re-described
Redescription mining as an unsupervised technique is different from subgroup discovery [12], [14], [19], [35] and its ability to find bi-directional associations distinguishes it from association rule mining [1], [15], [38] that provides unidirectional associations
In addition to providing thorough results of a framework’s performance and comparison to the naive way of creating multi-view redescription mining algorithms by utilizing existing two-view approaches, we provide experiments of using supplementing forest of Predictive Clustering trees (PCTs) [28], with [5] (PCT-ROS) and without [18] random output selection and using a supplementing forest of Extra multi-target Predictive Clustering trees [17] (EPCT) to increase the overall performance

Summary

Introduction

Redescription mining [30] aims to find multiple characterizations (re-descriptions) of different subsets of entities in a set of available data, i.e. to identify subsets of instances that can be re-described. These characterizations are expressed in a rule-like form which makes them easy to understand. The ability of redescription mining to provide tuples of rules re-describing entities in a support set (using multiple sources of data, used to form views) distinguishes this approach from related fields, e.g. clustering [8], [16], [36], conceptual clustering [7], [22] and multi-view clustering [2], [13]. Redescription mining as an unsupervised technique is different from subgroup discovery [12], [14], [19], [35] and its ability to find bi-directional associations distinguishes it from association rule mining [1], [15], [38] that provides unidirectional associations

Objectives

Methods

Results