The DOI matches its Crossref record; 33 of the 34 references it cites were found in a registry or an index; 1 of the 7 sources its claims cite were checked for support.
Improved Churn Causal Analysis Through Restrained High-Dimensional Feature Space Effects in Financial Institutions
Open accessAbstract
Abstract Customer churn describes terminating a relationship with a business or reducing customer engagement over a specific period. Customer acquisition cost can be five to six times that of customer retention, hence investing in customers with churn risk is wise. Causal analysis of the churn model can predict whether a customer will churn in the foreseeable future and identify effects and possible causes for churn. In general, this study presents a conceptual framework to discover the confounding features that correlate with independent variables and are causally related to those dependent variables that impact churn. We combine different algorithms including the SMOTE, ensemble ANN, and Bayesian networks to address churn prediction problems on a massive and high-dimensional finance data that is usually generated in financial institutions due to employing interval-based features used in Customer Relationship Management systems. The effects of the curse and blessing of dimensionality assessed by utilising the Recursive Feature Elimination method to overcome the high dimension feature space problem. Moreover, a causal discovery performed to find possible interpretation methods to describe cause probabilities that lead to customer churn. Evaluation metrics on validation data confirm the random forest and our ensemble ANN model, with %86 accuracy, outperformed other approaches. Causal analysis results confirm that some independent causal variables representing the level of super guarantee contribution, account growth, and account balance amount were identified as confounding variables that cause customer churn with a high degree of belief. This article provides a real-world customer churn analysis from current status inference to future directions in local superannuation funds.
Overview
CoChat Summary
TL;DR
This paper proposes a churn prediction and causal analysis framework for a local superannuation fund, combining RFE, SMOTE, ensemble ANNs, and Bayesian networks/DoWhy. It reports ensemble ANN and random forest reaching 86% test accuracy, with causal analysis identifying SG contribution recency, account growth, and account balance as churn-related confounders.
Key findings
- On 12-month observation and 6-month outcome windows, random forest achieved AUC = 80%; the proposed ensemble ANNs and RF had almost the same performance, improving AUC by 7.5% compared with logistic regression. In Table 2, RF and ensemble ANNs both reached test accuracy 0.86; RF had Cohen Kappa 0.51 and MCC 0.50, while ensemble ANNs had Cohen Kappa 0.45 and MCC 0.45.
- Causal analysis with propensity score-based inverse weighting estimated that high_sg_recency increased churn probability by ∼15% (estimate 0.156396), annualrpt_pref increased it by ∼14% (0.144440), stmt_pref decreased it by ∼14% (−0.142732), high_bal_change_amount decreased it by ∼12% (−0.123401), high_acc_balance decreased it by ∼9% (−0.091698), promotional_pref decreased it by ∼8% (−0.086401), low_account_growth increased it by ∼3% (−0.033853), and high_cust_tenure decreased it by ∼3% (−0.027969). Data Subset Refuter estimates were nearly identical.
- SHAP and PDP plots identified acc_balance_change_ratio, login_recency, acc_tenure, cust_tenure, and account_growth_change as robust causal estimands; promotion_pref and acc_tenure were among the most influential features pushing predictions higher, while acc_balance_change_amount and sg_recency pulled predictions lower.
- The conclusion states churn rate can be reduced by ∼3% for customers with active account > 1 year, and churn probability decreases by ∼9% when the customer has a high account balance over $100k.
Methods
- Data: 12 datasets from a local finance/superannuation company, each with approximately 250,000 examples and 88 features (71 numerical, 17 nominal). After feature selection, the study used 193 features from 124,363 total examples, with 80% training and 20% test sets; SMOTE equalized classes at 14,031 examples each.
- Models: compared the proposed ensemble ANNs against ten best-practice classifiers. The ensemble ANNs used two deep ANN architectures with 4 hidden layers, tanh and 3×relu activations, dropout 0.2/0.4, sigmoid output, ADAM optimization, 100 epochs, and batch size 512. RFE, SMOTE, majority-voting ensembles, SHAP, and PDP were also used.
- Causal analysis: Bayesian causal graphs and the DoWhy Python package were used with correlation analysis, propensity score-based inverse weighting, and the Data Subset Refuter to test assumptions.
Limitations
- The paper states that causal analysis has limitations in analysing the identified confounding effects with other popular causal inference methods like counterfactual analysis; future work should investigate smaller outcome windows and counterfactual causal analysis.
- The study is based on data from one local superannuation fund, and the dataset was unidentified by the company; no external validation is reported.
- Evident from the text: causal conclusions rely on assumptions encoded in a Bayesian causal graph and observational data, so the analysis supports high-belief confounding relationships but does not establish counterfactual causality.
Model-written summary. Check it against the paper before relying on it.
Integrity signals
Registry
Cites
Statistics
Verification
Claims the paper makes
1 Valid · 5 Not valid · 5 Not checked · 31 Own results and interpretations
Every source the paper cites for this claim was checked and supports it.
In the paper
“Moreover, the RemsProp training approach outperformed”
What the sources say
- Supportedcites [1] Domingos BO, Daramola O. Experimental analysis of hyperparameters for deep learning-based churn prediction in the banking sector. Computation. 2021;9(3):34. source
Section 4.2 (page not supplied) reports that RMSProp gave the highest accuracy for both the MLP (84.5%) and DNN (86.45%) in the optimizer comparison. The methods passage lists SGD, Adam, AdaGrad, Adadelta, and AdaMax among the algorithms tested. The claim's spelling, “RemsProp,” refers to the paper's RMSProp. (LLM, conf=0.92)
plainstates a cause- Asserted: passes — The authors put it forward as true.
- Checkable: passes — It can be checked against evidence.
- Load-bearing: passes — The paper’s conclusions depend on it.
- Based: passes — The paper gives a reason why it holds.
- Supported
A result the paper reports from its own data — it cites no other source for it, so there is none to check. The quote shows it is in the paper.
In the paper
“The RF outperformed the other algorithms with AUC = 80%, whereas the proposed ensemble ANNs and RF have almost the same performance, with improved AUC by 7.5% compared with logistic regression.”
plainstates a cause- Asserted: passes — The authors put it forward as true.
- Checkable: passes — It can be checked against evidence.
- Load-bearing: passes — The paper’s conclusions depend on it.
- Based: passes — The paper gives a reason why it holds.
A result the paper reports from its own data — it cites no other source for it, so there is none to check. The quote shows it is in the paper.
In the paper
“For instance, Linear regression estimation indicates that estimated effect = - 0.033853 corresponds to churn probability reducing by ≈ 3% when the customer has lower account growth rate.”
hedgedstates a cause- Asserted: passes — The authors put it forward as true.
- Checkable: passes — It can be checked against evidence.
- Load-bearing: passes — The paper’s conclusions depend on it.
- Based: passes — The paper gives a reason why it holds.
- Rests on 2 assumptions not shown here.
A result the paper reports from its own data — it cites no other source for it, so there is none to check. The quote shows it is in the paper.
In the paper
“Figure 4 visualises customer and account tenure distribution data that are almost identical in all datasets.”
plain- Asserted: passes — The authors put it forward as true.
- Checkable: passes — It can be checked against evidence.
- Load-bearing: passes — The paper’s conclusions depend on it.
- Based: passes — The paper gives a reason why it holds.
A result the paper reports from its own data — it cites no other source for it, so there is none to check. The quote shows it is in the paper.
In the paper
“Evaluation metrics on validation data confirm the random forest and our ensemble ANN model, with %86 accuracy, outperformed other approaches.”
plainstates an association- Asserted: passes — The authors put it forward as true.
- Checkable: passes — It can be checked against evidence.
- Load-bearing: passes — The paper’s conclusions depend on it.
- Based: passes — The paper gives a reason why it holds.
18 more claims
Create a free account to see all 42 claims and every source verdict.
- It may state two separate claims as one. 9
- Probably not a substantial claim. 13
Self-check
Percentages, sums and reported differences re-computed from the paper’s own numbers.
Each contradiction between two passages listed here was confirmed by an independent check.
Rated major: it would touch a result — a value a result rests on, reported and recomputed p-values on opposite sides of the significance level, two passages that conflict with no sign which is right, or a cited work retracted before this paper was published.
In the paper
§ 4.3 Churn Propensity Modelling
“After feature selection, the proposed algorithms were employed for training (80%), and test (20%) sets to derive 193 features from 124,363 total examples.”
§ 4.3 Churn Propensity Modelling
“we employed the synthetic minority oversampling technique (SMOTE) [17] in pre-processing to synthesise new examples for the minority class to ensure equal sample count (14,031) for each class (1 or 0) in the training set.”
Rated major: it would touch a result — a value a result rests on, reported and recomputed p-values on opposite sides of the significance level, two passages that conflict with no sign which is right, or a cited work retracted before this paper was published.
In the paper
§ 6.2 Causality Analysis Results
“estimated effect = - 0.033853 corresponds to churn probability reducing by ≈ 3% when the customer has lower account growth rate.”
§ 6.2 Causality Analysis Results
“| low_account_growth | - 0.033853 | - 0.033920 | increased by ∼ 3% |”
Rated minor: it would not touch a result — a value no result rests on, a p-value that stays on the same side of the significance level, or a conflict that is clerical, resolved elsewhere in the paper, or outside its own findings.
In the paper
§ Abstract
“Evaluation metrics on validation data confirm the random forest and our ensemble ANN model, with %86 accuracy, outperformed other approaches.”
§ 6.1 Prediction Results
“| ExtraTrees | 0.86 | 0.49 | 0.49 |”
Rated minor: it would not touch a result — a value no result rests on, a p-value that stays on the same side of the significance level, or a conflict that is clerical, resolved elsewhere in the paper, or outside its own findings.
In the paper
§ Abbreviations
“| DAG | Specific magnetisation |”
§ 2.2 Studies on Causal Inference of Churn
“Directed acyclic graph (DAG) has been proposed to represent causal relationships in Bayesian networks [10].”
Rated minor: it would not touch a result — a value no result rests on, a p-value that stays on the same side of the significance level, or a conflict that is clerical, resolved elsewhere in the paper, or outside its own findings.
In the paper
§ 4.3 Churn Propensity Modelling
“We compared the proposed DFF NNs algorithm approach with seven current best-practise classifiers”
§ 7 Conclusion
“accuracy was compared between the proposed ensemble ANNs and ten best-practise classifiers.”
2 more findings
Create a free account to see every finding of this check.
Citations
Cites
SMOTE synthetic minority over-sampling technique
Chawla, N. V., Bowyer, K. W., Hall, L. O., & Kegelmeyer, W. P. (2002) · Journal of Artificial Intelligence Research · doi:10.1613/jair.953 · Cited by 32,865
The central role of the propensity score in observational studies for causal effects
ROSENBAUM, P. R., & RUBIN, D. B. (1983) · Biometrika · doi:10.1093/biomet/70.1.41 · Cited by 31,241
An Introduction to Statistical Learning
James, G., Witten, D., Hastie, T., & Tibshirani, R. (2013) · Springer Texts in Statistics · doi:10.1007/978-1-4614-7138-7 · Cited by 12,106
Gene selection for cancer classification using support vector machines
Guyon, I., Weston, J., Barnhill, S., & Vapnik, V. (2002) · Machine Learning · doi:10.1023/a:1012487302797 · Cited by 9,954
Wrappers for feature subset selection
Kohavi, R., & John, G. H. (1997) · Artificial Intelligence · doi:10.1016/s0004-3702(97)00043-x · Cited by 9,102
The use of the area under the ROC curve in the evaluation of machine learning algorithms
Bradley, A. P. (1997) · Pattern Recognition · doi:10.1016/s0031-3203(96)00142-2 · Cited by 7,217
The Central Role of the Propensity Score in Observational Studies for Causal Effects
Rosenbaum, P. R., & Rubin, D. B. (1983) · Biometrika · doi:10.2307/2335942 · Cited by 3,495
Causal inference in statistics: An overview
Pearl, J. (2009) · Statistics Surveys · doi:10.1214/09-ss057 · Cited by 2,336 · cited 2×
“Recent causal inference developments highlight fundamental alterations required to move from standard statistical analysis toward multivariate causal analysis [3].”
BackgroundNeural Networks and Deep Learning
Theodoridis, S. (2015) · Elsevier eBooks · doi:10.1016/b978-0-12-801522-3.00018-5 · Cited by 942
Ullah, I., Raza, B., Malik, A. K., Imran, M., Islam, S. U., & Kim, S. W. (2019) · IEEE Access · doi:10.1109/access.2019.2914999 · Cited by 420
Research works printed but not linked: 2
- Karp AH. Using logistic regression to predict customer retention. In: Proceedings of the eleventh northeast SAS users group conference. 1998. http://www.lexjansen.com/nesug/nesug98/solu/p095.pdf.
- Donoho DL. High-dimensional data analysis: the curses and blessings of dimensionality. AMS Math Chall Lecture. 2000;1(2000):32.
Cited by
Manzoor, A., Qureshi, M. A., Kidney, E., & Longo, L. (2024) · IEEE Access · doi:10.1109/access.2024.3402092 · Cited by 192
“Hason et al. [145] employed ensemble ANN with the Bayesian network and compared their results with several other tree-based classifiers, concluding that random forest and ensemble classifiers outperform the others.”
BackgroundResultPredict customer churn using combination deep learning networks model
Vu, V.-H. (2023) · Neural Computing and Applications · doi:10.1007/s00521-023-09327-w · Cited by 24
Gottipati, K. C., & Maddineni, D. (2025) · doi:10.1109/icdsis65355.2025.11070746 · Cited by 3
“Rudd et al. [18] utilized a combination of ML methods, such as Bayesian networks, SMOTE, and ensemble ANN, to address the high-dimensional finance data for churn prediction.”
IMPROVED DOUBLE-LEVEL BLENDING MODEL (IDLBM): CUSTOMER CHURN ESTIMATING IN INSURANCE INDUSTRY
Jajam, N., Challa, N. P., & Prasanna, K. L. (2023) · doi:10.1109/icaect57570.2023.10118051 · Cited by 2
“This is essentially a classification problem, in that each customer must be classified as a potential churner or a non-churner [2].”
BackgroundKernelized Extreme Learning Machine Enabled Churn Predictive Financial Risk Assessment Model
Dimlo, U. M. F., Huerta-Soto, R., Vargas, L. N., Jiménez, J. T., Reyes, C. A. R., & Girdharwal, N. (2022) · International Conference on Augmented Intelligence and Sustainable Systems (ICAISS) · doi:10.1109/icaiss55157.2022.10010834 · Cited by 2
Metadata
- Authors
- Hason Rudd, D., Huo, H., & Xu, G.
- Affiliations
- University of Technology Sydney (AU)
- Venue
- Human-Centric Intelligent Systems
- Published
- 2022-07-27
- Volume
- 2
- Issue
- 3-4
- Pages
- 70-80
- ISSN
- 2667-1336 (ISSN-L)
- OpenAlex
- W4288066831
- Cited by
- 12
- References
- 34
- Citations by year
- 2026: 1 · 2025: 4 · 2024: 2 · 2023: 4 · 2022: 1
- FWCI
- 2.20
- Type
- journal article
- Language
- en
- Open access
- diamond
- doi.org
- Licence
- cc-by
- creativecommons.org/licenses/by/4.0 (tdm)
- creativecommons.org/licenses/by/4.0 (vor)
- Locations
- Human-Centric Intelligent Systems (publishedVersion, cc-by)
- arXiv (Cornell University) (submittedVersion, cc-by)
- HAL (submittedVersion, public-domain)
- DOAJ (DOAJ: Directory of Open Access Journals) (submittedVersion)
- arXiv (Cornell University) (cc-by)
- Funders
- Australian Research Council (DP22010371,LE220100078, DP22010371, DP200101374, LP170100891, LE220100078)
- Topics
- Social Sciences › Business, Management and Accounting › Marketing › Customer churn and segmentation
- Social Sciences › Business, Management and Accounting › Marketing › Customer Service Quality and Loyalty
- Physical Sciences › Computer Science › Information Systems › Data Mining and Machine Learning Applications
- Keywords
- customer churn, causal discovery, confounding variables, high-dimensional feature space, curse of dimensionality, SMOTE, neural network ensembles, Bayesian networks, recursive feature elimination, customer relationship management, superannuation funds, financial institutions, customer retention, account balances
About these checks
Registry, retraction, Cites and statistics results are computed, not judged, with one exception: a printed reference that no index lists is read once by a model before it counts as a citation error. Claims are extracted by a model and kept only when their quote appears verbatim in the text; each cited source is then checked, with a model’s help, for whether it says what the claim relies on. Those verdicts can be wrong — read the claim’s quote and the cited source before relying on one.
A language model drew the claims from the paper on 2026-09-29, under version 3 of the claim definition (model deepseek-v4.1-flash, prompt extract-v3).
A second model labelled each claim on 2026-09-29 (model jev-1.13-20260917, definition revision gold-guideline-v0.2).
Cited sources are checked with a language model’s help (gpt-6-sol).
Statistics and arithmetic are recomputed, not judged.
Confidence bands are shown only for fields calibrated on our annotated test set. None is calibrated yet, so no band is shown.
What counts as a claim ›How we measure confidence ›How we check this ›
Are you an author, or do you see a problem with this page? Request a review
How to cite
Cite CoChat for the checks on this page, and the paper for its own findings.
- This page
- CoChat. (2026). Citation check: Improved Churn Causal Analysis Through Restrained High-Dimensional Feature Space Effects in Financial Institutions. https://app.cochat.ai/papers/10.1007/s44230-022-00006-y
- The paper
- Hason Rudd, D., Huo, H., & Xu, G. (2022). Improved Churn Causal Analysis Through Restrained High-Dimensional Feature Space Effects in Financial Institutions. Human-Centric Intelligent Systems. https://doi.org/10.1007/s44230-022-00006-y
Data sources
Bibliographic data: Semantic Scholar, OpenAlex, Crossref.
Contains information from the Semantic Scholar Academic Graph Datasets, made available under the ODC Attribution License (ODC-BY 1.0).
Semantic Scholar’s data is described in: Kinney, R., et al. (2023). The Semantic Scholar Open Data Platform. arXiv:2301.10140.