Search Conferences

Type in any word, words or author name. This searchs through the abstract title, keywords and abstract text and authors. You may search all conferences or just select one conference.

 All Conferences
 EMAC 2019 Annual Conference
 EMAC 2020 Annual Conference
 EMAC 2020 Regional Conference
 EMAC 2021 Annual Conference
 EMAC 2021 Regional Conference
 EMAC 2022 Annual
 EMAC 2022 Regional Conference

EMAC 2019 Annual Conference

A Framework of Unsupervised Machine Learning Algorithms for User Profiling

Published: May 28, 2019


Erik Kuiper, University of Twente; Efthymios Constantinides, University of Twente; Sjoerd de Vries, University of Twente; Robert Marinescu-Muster, University of Twente; Floris Metzner, University of Twente


Unsupervised Machine Learning ; Data-Driven Segmentation ; Digital Marketing


Organizations often have difficulties to extract knowledge from data and selecting appropriate Machine Learning algorithms in order to develop accurate Behavioural Profiles or user segments. Moreover, marketing departments often lack a fundamental understanding on data-driven segmentation methodologies. This paper aims to develop a framework outlining Unsupervised Machine Learning algorithms for the purpose of User Profiling with respect to important data properties. A systematic literature review was conducted on the most prominent Unsupervised Machine Learning algorithms and their requirements regarding the characteristics of the dataset. A framework is proposed outlining various Unsupervised Machine Learning algorithms for User Profiling. It provides two-stage clustering strategies for categorical, numerical, and mixed types of data with respect to the data size and data dimensionality. The first stage consists of an hierarchical or model-based clustering algorithm to determine the number of clusters. In the second stage, a non-hierarchical clustering algorithm is applied for cluster refinement. The framework can support researchers and practitioners to determine which Unsupervised Machine Learning algorithms are appropriate for developing robust behavioural profiles or data-driven user segments.