Full Text Available
Note: Clicking the button above will open the full text document at the original institutional repository in a new window.
A framework for modelling customer invoice payment predictions
Moore, W. R. 2025. A framework for modelling customer invoice payment predictions. Unpublished doctoral dissertation. Stellenbosch: Stellenbosch University [online]. Available: https://scholar.sun.ac.za/items/42773281-13ed-452b-9d09-1cfebefc123a
Saved in:
| Main Author: | |
|---|---|
| Other Authors: | |
| Format: | Thesis |
| Language: | English |
| Published: |
Stellenbosch : Stellenbosch University
2025
|
| Subjects: | |
| Tags: |
No Tags, Be the first to tag this record!
|
| _version_ | 1867613846014787584 |
|---|---|
| access_status_str | Open Access |
| author | Moore, Willem Roux |
| author2 | Van Vuuren, J. H. |
| author_browse | Moore, Willem Roux Van Vuuren, J. H. |
| author_facet | Van Vuuren, J. H. Moore, Willem Roux |
| author_sort | Moore, Willem Roux |
| collection | Thesis |
| dc_rights_str_mv | Stellenbosch University |
| description | Moore, W. R. 2025. A framework for modelling customer invoice payment predictions. Unpublished doctoral dissertation. Stellenbosch: Stellenbosch University [online]. Available: https://scholar.sun.ac.za/items/42773281-13ed-452b-9d09-1cfebefc123a |
| format | Thesis |
| id | oai:scholar.sun.ac.za:10019.1/132588 |
| institution | Stellenbosch University (South Africa) |
| language | English |
| last_indexed | 2026-06-10T12:42:37.450Z |
| license_str | Other — see source repository |
| provenance_str_mv | Harvested via OAI-PMH from SUNScholar — Stellenbosch University Repository |
| publishDate | 2025 |
| publishDateRange | 2025 |
| publishDateSort | 2025 |
| publisher | Stellenbosch : Stellenbosch University |
| publisherStr | Stellenbosch : Stellenbosch University |
| record_format | dspace |
| source_str | SUNScholar — Stellenbosch University Repository |
| spelling | oai:scholar.sun.ac.za:10019.1/132588 A framework for modelling customer invoice payment predictions Moore, Willem Roux Van Vuuren, J. H. Stellenbosch University. Faculty of Engineering. Dept. of Industrial Engineering. Accounts receivable -- Data processing Credit scoring systems Credit -- Mathematical models UCTD Moore, W. R. 2025. A framework for modelling customer invoice payment predictions. Unpublished doctoral dissertation. Stellenbosch: Stellenbosch University [online]. Available: https://scholar.sun.ac.za/items/42773281-13ed-452b-9d09-1cfebefc123a Thesis (PhD)--Stellenbosch University, 2025. ENGLISH ABSTRACT: Most companies offer credit sales to their customers, thereby providing customers with a financial service as well as other services or products. The act of granting credit, however, incurs the cost of money tied up in accounts receivable and increased administration costs incurred to collect outstanding balances. The collection of accounts receivable is eminently important to any business in pursuit of a healthy cashflow position. The management of credit sales is, however, performed manually in most businesses, making it time-consuming, expensive, and inaccurate. This situation has given rise to a relatively new research field of invoice payment prediction, which involves proactively attempting to identify those invoices from accounts receivable that are likely to be paid late, or not at all. Several quantitative techniques which stem from the academic literature on credit scoring have been applied in this field of research, particularly techniques from the realms of machine learning and survival analysis. While an abundance of research has been dedicated to the design and evaluation of algorithms for predicting invoice payment dates, little guidance is available in the literature on the data preparation phase of the model development process. This is a particularly relevant shortcoming, since most of the existing modelling approaches towards invoice payment prediction have been applied in the context where payments are linked to specific invoices, while many companies adopt a risk-averse approach according to which payments are allocated to the oldest debt first. Moreover, most existing modelling approaches focus on predicting whether an invoice will be paid, with few considering the problem of when an invoice will be paid. Another gap in the literature pertains to the ensembling of survival analytic techniques in a sequential fashion. In this dissertation, a generic framework for modelling customer invoice payment predictions is proposed. The aim of the framework is to facilitate the entire process of preparing transaction data for the purposes of further analysis, generating relevant features from past customer behaviours, and selecting and ensembling suitable models for predicting the times to payment associated with invoices. A new sequential ensembling approach, called the Survival Boost algorithm, is also introduced and the improvement it brings about in terms of model performance is demonstrated. The framework has been designed to support future studies in transforming raw transaction data into a format suitable for modelling and it provides a sound foundation from which future research may ensue. The aforementioned framework is verified by applying a computerised instantiation of the framework to a publicly available data set. A similar implementation is also applied in the context of a real-world case study, so as to demonstrate the practical value added by the framework. Moreover, the potential improvement in predictive performance when invoking the Survival Boost algorithm is illustrated in both problem instances, thereby proving the algorithm’s ability to generalise across different problem contexts. The framework proposed in this dissertation is therefore shown to address successfully and overcome the shortcomings of similar frameworks in the literature. AFRIKAANSE OPSOMMING: Die meeste maatskappye bied kredietverkope aan hul kli¨ente en verskaf sodoende aan kli¨ente ’n finansi¨ele diens sowel as ander dienste of produkte. Die aanbod van krediet bring ’n koste mee as gevolg van geld wat vasgemaak is in rekeninge ontvangbaar asook verhoogde administrasiekoste wat aangegaan word om uitstaande saldo’s in te vorder. Die invordering van rekeninge ontvangbaar is uiters belangrik vir enige besigheid in die nastrewing van ’n gesonde kontantvloeiposisie. Die bestuur van kredietverkope word egter in die meeste besighede met die hand uitgevoer, wat dit tydrowend, duur en onakkuraat maak. Hierdie situasie het aanleiding gegee tot ’n relatief nuwe navorsingsveld van voorspelling van kli¨ente-faktuurbetalings, wat behels dat daar pro-aktief gepoog word om d´a´ardie fakture uit rekeninge ontvangbaar te identifiseer wat waarskynlik laat, of glad nie, betaal sal word. Verskeie kwantitatiewe tegnieke wat uit die akademiese literatuur oor krediet-evaluering spruit, is al in hierdie navorsingsveld toegepas, veral tegnieke uit die gebiede van masjienleer en oorlewingsanalise. Terwyl ’n oorvloed navorsing reeds gewy is aan die ontwerp en evaluering van algoritmes vir die voorspelling van faktuurbetalingsdatums, is daar min riglyne in die literatuur beskikbaar oor die datavoorbereidingsfase van die modelontwikkelingsproses. Dit is ’n besonder relevante tekortkoming, aangesien die meeste van die bestaande modelleringbenaderings tot faktuurbetalingsvoorspelling toegepas is in die konteks waar betalings aan spesifiekefakture gekoppel word, terwyl baie maatskappye ’n risiko-vermydende benadering volg waarvolgens betalings aan die oudste skuld eerste toegewys word. Boonop fokus die meeste bestaande modelleringsbenaderings daarop om te voorspel of ’n faktuur betaal sal word, met min wat die probleem van wanneer ’n faktuur betaal sal word, oorweeg. Nog ’n leemte in die literatuur het betrekking op die samestelling van oorlewingsanalitiese tegnieke op ’n opeenvolgende wyse. In hierdie proefskrif word ’n generiese raamwerk vir die modellering van kli¨ente-faktuurbetalingsvoorspellings voorgestel. Die raamwerk het ten doel om die hele proses van voorbereiding van transaksiedata vir die doeleindes van verdere analise te fasiliteer, relevante kenmerke uit vorige kli¨entegedrag te genereer, en geskikte modelle te kies en saam te stel vir die voorspelling van die tye tot betaling met fakture geassosieer. ’n Nuwe opeenvolgende samestellingsbenadering, bekend as die Survival Boost-algoritme, word ook daargestel en die verbetering wat dit in terme van modelprestasie teweegbring, word gedemonstreer. Die raamwerk is ontwerp om toekomstige studies te ondersteun in die transformasie van rou transaksiedata in ’n formaat wat geskik is vir modellering en dit bied ’n goeie grondslag waaruit toekomstige navorsing kan voortspruit. Die bogenoemde raamwerk word geverifieer deur ’n gerekenariseerde instansiasie daarvan op ’n publiek-beskikbare datastel toe te pas. ’n Soortgelyke implementasie word ook in die konteks van ’n werklike gevallestudie toegepas om die praktiese waarde wat deur die raamwerk toegevoeg word, te demonstreer. Boonop word die potensi¨ele verbetering in voorspellende prestasie wanneer die Survival Boost-algoritme gebruik word, in beide probleemgevalle ge¨ıllustreer, en sodoende word die algoritme se vermo¨e om oor verskillende probleemkontekste te veralgemeen, bewys. Daar word dus getoon dat die raamwerk wat in hierdie proefskrif voorgestel word, daartoe in staat is om die tekortkominge van soortgelyke raamwerke in die literatuur suksesvol aan te spreek en te oorkom. Doctoral 2025-06-11T10:26:35Z 2025-06-11T10:26:35Z 2025-03 Thesis https://scholar.sun.ac.za/handle/10019.1/132588 en Stellenbosch University xxii, 195 pages : illustrations application/pdf Stellenbosch : Stellenbosch University |
| spellingShingle | Accounts receivable -- Data processing Credit scoring systems Credit -- Mathematical models UCTD Moore, Willem Roux A framework for modelling customer invoice payment predictions |
| title | A framework for modelling customer invoice payment predictions |
| title_full | A framework for modelling customer invoice payment predictions |
| title_fullStr | A framework for modelling customer invoice payment predictions |
| title_full_unstemmed | A framework for modelling customer invoice payment predictions |
| title_short | A framework for modelling customer invoice payment predictions |
| title_sort | framework for modelling customer invoice payment predictions |
| topic | Accounts receivable -- Data processing Credit scoring systems Credit -- Mathematical models UCTD |
| url | https://scholar.sun.ac.za/handle/10019.1/132588 |
| work_keys_str_mv | AT moorewillemroux aframeworkformodellingcustomerinvoicepaymentpredictions AT moorewillemroux frameworkformodellingcustomerinvoicepaymentpredictions |