Papp, István; (2023) Supervised Preference Models: Data and Storage, Methods, and Tools for Application. Masters thesis (M.Phil), UCL (University College London).

Preview

Text Istvan Papp MPhil thesis post corrections.pdf - Accepted Version Download (2MB) | Preview

Abstract

In this thesis, we present a variety of models commonly known as pairwise comparisons, discrete choice and learning to rank under one paradigm that we call preference models. We discuss these approaches together with the intention to show that these belong to the same family and show a unified notation to express these. We focus on supervised machine learning approaches to predict preferences, present existing approaches and identify gaps in the literature. We discuss reduction and aggregation, a key technique used in this field and identify that there are no existing guidelines for how to create probabilistic aggregations, which is a topic we begin exploring. We also identify that there are no machine learning interfaces in Python that can account well for hosting a variety of types of preference models and giving a seamless user experience when it comes to using commonly recurring concepts in preference models, specifically, reduction, aggregation and compositions of sequential decision making. Therefore, we present our idea of what such software should look like in Python and show the current state of the development of this package which we call skpref.

Download activity - last month

Export as CSVJSONXML

Download activity - last 12 months

Export as JSONXMLCSV

Downloads by country - last 12 months

Export as CSVXMLJSON

Archive Staff Only

View Item