CEU Electronic Theses and Dissertations, 2019
| Author | Tóth, Bence László |
|---|---|
| Title | Standard Warranty Data Mart & Predictive Modeling Of Potential Open Items |
| Summary | The Commercial team in GE Healthcare struggled with the current reconciliation process for the Standard Warranty accounts. It was a highly manual, non-standardized, prone to error activity and required constant manual effort to monitor the state of the accounts including open items occasionally appearing which need action from the team. Furthermore, it did not support the reporting requirements for proper internal auditing. After a brief summary of accounting terms and problem statement, the Capstone project offers a solution by the development of an automated wing-to-wing reconciliation: daily automatic ETL scripts consolidate batch data from BOXi, OBIEE, and manual sources. These are stored in a Microsoft SQL Server environment, where triggers and stored procedures create a detailed, order-level view on reconciliation status. This automation and simplification has an important effect on quality and saves around 21 hours for the team while providing daily visibility on movements, instead of the once-per-quarter approach of the previous process. From the reconciliation view, an R script trains an XGBoost model on the data and predicted probabilities are assigned to the binary classes based on the sample mean threshold of the occurrence of open items within the data, resulting in an overall accuracy of 94% in differentiating them from recorded items. |
| Supervisor | Sallo, Laszlo |
| Department | Economics MSc |
| Full text | https://www.etd.ceu.edu/2019/toth_bence02.pdf |
Visit the CEU Library.
© 2007-2021, Central European University