A Machine Learning Framework for De-Risking ERP Changes and Ensuring Financial Reporting Stability

View at publisher Summarise with VIRAT
Authors: Puneet Thakkar, Swetha Chinta
Cited by: 0
References: 13

Summary

No summary yet. VIRAT can read this paper and write a plain-language summary of its aim, methods, key findings and limitations.

Summarise with VIRAT Free to try with your Science Catalogs account.

Abstract

ERP systems underpin financial operations in contemporary organizations, and even minor system changes can seriously compromise the accuracy of financial reporting. Conventional approaches to managing changes are highly reliant on subjective judgment and tend to not provide quantification of risk. This paper introduces a machine learning framework for predicting the impact that ERP change requests may have on financial reporting by using an 8,050-record dataset compiled across multiple implementations. Fifteen technical, procedural, and temporal features were engineered, standardized, and tested across four models: Support Vector Machine (SVM), Random Forest, Gradient Boosting, and Logistic Regression. Using stratified 10-fold cross-validation, the SVM outperformed other contenders with 99.19% accuracy, 98.9 % precision, and 99.3 % recall. Feature importance analysis identified Priority Level and Data Model Changes as the strong leading predictors of stability outcomes. The proposed framework establishes a quantitative and reproducible means of assessing the risk of ERP changes through a reduction in the level of system instability and improvement in the reliability of financial reporting. This work allows one to set a benchmark for data-driven ERP governance and underlines prospects for the real-time prediction of impact, platform-specific modeling, and integration into CI/CD-based workflows of change management.

Fields of Study & Subjects

ERP Systems Implementation and Impact Robotic Process Automation Applications Decision Support System Applications

Publication details

Journal / source
International Conference on Communication, Computing and Electronics Systems (ICCCES)
Pages
1229–1235
Publication date
21 January 2026
Language
EN