Full Text Available
Note: Clicking the button above will open the full text document at the original institutional repository in a new window.
System dynamics and machine learning hybrid model for churn prediction in retail banking
Dissertation (MEng (Industrial Engineering))--University of Pretoria, 2026.
Saved in:
| Other Authors: | |
|---|---|
| Format: | Thesis |
| Language: | English |
| Published: |
University of Pretoria
2026
|
| Subjects: | |
| Tags: |
No Tags, Be the first to tag this record!
|
| _version_ | 1869483849920020480 |
|---|---|
| access_status_str | Open Access |
| author2 | Ayomoh, Michael |
| author_browse | Ayomoh, Michael |
| author_facet | Ayomoh, Michael |
| collection | Thesis |
| dc_rights_str_mv | © 2024 University of Pretoria. All rights reserved. The copyright in this work vests in the University of Pretoria. No part of this work may be reproduced or transmitted in any form or by any means, without the prior written permission of the University of Pretoria. |
| description | Dissertation (MEng (Industrial Engineering))--University of Pretoria, 2026. |
| format | Thesis |
| id | oai:repository.up.ac.za:2263/108410 |
| institution | University of Pretoria (South Africa) |
| language | English |
| last_indexed | 2026-07-01T04:05:32.605Z |
| license_str | Other — see source repository |
| provenance_str_mv | Harvested via OAI-PMH from UPSpace — University of Pretoria Institutional Repository |
| publishDate | 2026 |
| publishDateRange | 2026 |
| publishDateSort | 2026 |
| publisher | University of Pretoria |
| publisherStr | University of Pretoria |
| record_format | dspace |
| source_str | UPSpace — University of Pretoria Institutional Repository |
| spelling | oai:repository.up.ac.za:2263/108410 System dynamics and machine learning hybrid model for churn prediction in retail banking Ayomoh, Michael u18181652@tuks.co.za Monyemonwa, Kakgiso Vanessa UCTD Sustainable Development Goals (SDGs) Machine Learning System Dynamics Retail banking Churn Predictive modelling Dissertation (MEng (Industrial Engineering))--University of Pretoria, 2026. Customer churn in retail banking represents a dynamic, feedback-driven system in which customer satisfaction, service quality, acquisition, retention, and marketing interactions evolve over time. Existing churn prediction models predominantly rely on static machine learning approaches that fail to capture these endogenous churn dynamics. Resulting in models that have limited explanatory insight. This study develops and evaluates an integrated System Dynamics–Machine Learning (SD–ML) modelling framework to explain, simulate, and predict customer churn in a retail banking context. A system dynamics model was first constructed to represent the structural feedback mechanisms governing customer acquisition, retention, and attrition. The SD model puts into practice and tests the effects of service quality, customer satisfaction, marketing effectiveness, and word-of-mouth. At its core the SD model was used to mirror the complex, dynamic nature of customer churn. Quantify the interconnectedness and interrelationships of key churn variables within the system of churn. The SD model was validated and used to generate behavioural stock-and-flow trajectories, which are then extracted as dynamic features. These SD-generated features are integrated with source data, which included the demographic, financial, and campaign data. The dataset pipeline used to train and validate supervised machine learning classifiers entailed the SD-generated features and the preprocessed source dataset. The data preprocessing process included advanced feature engineering and feature selection. This was done extensively to enhance the predictive ability of the hybrid SD-ML model. The ML algorithms used in this study included XGBoost, CatBoost, and Random Forest. A total of three SD-ML models were built and compared using the standard metrics and the best-performing model. XGBoost was found to be the best-performing model based on the standard metrics, and was used to extract a forecast of customer churn, important feature variables and customer segmentation. Results demonstrate that incorporating SD-derived behavioural features improves churn trajectory forecasting and supports continuous customer segmentation based on churn risk and engagement patterns. The hybrid SD–ML model employing the XGBoost method demonstrated exceptional performance, with an accuracy of 99.80%, precision of 96.71%, recall of 97.05%, and an F1-score of 99.81%. An examination of feature importance showed that financial behaviour traits were the most important factors in predicting customer churn, accounting for 87% of the total predictive power. Customer segmentation revealed that segment one, which includes the young, low-income customers with significant credit risk and minimal subscription uptake, is the segment with the highest churn risk. This SD-ML hybrid uniquely models churn as a dynamic, feedback-driven system that combines the explanatory system insights with the data-driven prediction. The model provides an engineering decision-support tool for proactive churn detection, scenario testing, and targeted retention strategy design within complex service systems. Industrial and Systems Engineering MEng (Industrial Engineering) Restricted Faculty of Engineering, Built Environment and Information Technology SDG-08: Decent work and economic growth 2026-02-18T13:14:17Z 2026-02-18T13:14:17Z 2026-04-14 2026-01-28 Dissertation * A2026 http://hdl.handle.net/2263/108410 Disclaimer Letter en © 2024 University of Pretoria. All rights reserved. The copyright in this work vests in the University of Pretoria. No part of this work may be reproduced or transmitted in any form or by any means, without the prior written permission of the University of Pretoria. application/pdf University of Pretoria |
| spellingShingle | UCTD Sustainable Development Goals (SDGs) Machine Learning System Dynamics Retail banking Churn Predictive modelling System dynamics and machine learning hybrid model for churn prediction in retail banking |
| title | System dynamics and machine learning hybrid model for churn prediction in retail banking |
| title_full | System dynamics and machine learning hybrid model for churn prediction in retail banking |
| title_fullStr | System dynamics and machine learning hybrid model for churn prediction in retail banking |
| title_full_unstemmed | System dynamics and machine learning hybrid model for churn prediction in retail banking |
| title_short | System dynamics and machine learning hybrid model for churn prediction in retail banking |
| title_sort | system dynamics and machine learning hybrid model for churn prediction in retail banking |
| topic | UCTD Sustainable Development Goals (SDGs) Machine Learning System Dynamics Retail banking Churn Predictive modelling |
| url | http://hdl.handle.net/2263/108410 |