Skip to main content
Book cover

The Digital Journey of Banking and Insurance, Volume I pp 67–79Cite as

Palgrave Macmillan

AI for Impairment Accounting

  • 424 Accesses

Abstract

The authors present a practical application of machine learning in the context of an accounting department. The article gives an insight into how the use case was identified and how it was embedded in the existing IT landscape and infrastructure. The results of the chosen approach are presented, and an outlook is given at the end of the article.

Keywords

  • Impairment IFRS 9
  • Batch processing
  • Pattern recognition
  • Process monitoring
  • Anomaly detection
  • Data quality
  • Deep learning
  • Machine learning
  • Autoencoder
  • Batch
  • Bayesian network
  • H2O

This is a preview of subscription content, access via your institution.

Chapter
USD   29.95
Price excludes VAT (USA)
  • DOI: 10.1007/978-3-030-78814-8_6
  • Chapter length: 13 pages
  • Instant PDF download
  • Readable on all devices
  • Own it forever
  • Exclusive offer for individuals only
  • Tax calculation will be finalised during checkout
eBook
USD   69.99
Price excludes VAT (USA)
  • ISBN: 978-3-030-78814-8
  • Instant PDF download
  • Readable on all devices
  • Own it forever
  • Exclusive offer for individuals only
  • Tax calculation will be finalised during checkout
Hardcover Book
USD   89.99
Price excludes VAT (USA)
Learn about institutional subscriptions
Fig. 1

(© Helaba)

Fig. 2

(© Helaba)

Fig. 3

(© Helaba)

Fig. 4

(© Helaba)

Fig. 5

(© Helaba)

Fig. 6

(© Helaba)

Fig. 7

(© Helaba)

Fig. 8

(© Helaba)

Fig. 9

(© Helaba)

Notes

  1. 1.

    GAFAM (Google, Amazon, Facebook, Apple and Microsoft) and BATX (Baidu, Alibaba, Tencent and Xiaomi).

  2. 2.

    The dysfunctional cost cutting would be to brute force reduction of needed functionality.

  3. 3.

    The Proof of Concept (POC) only took eight weeks.

  4. 4.

    It was a big advantage that the POC could use the ICS rules as labels.

  5. 5.

    The impairment system calculates the impairment value by using credit risk methods like life-time expected loss for stage 2 and 3 transactions.

  6. 6.

    See information about the R package (Wickham et al. 2020) and, for an introduction, see Institute for Statistics and Mathematics of WU (2020).

  7. 7.

    Deep Feed Forward network—see Section 3.1 in Liermann et al., Deep Learning—An Introduction (2019a).

  8. 8.

    See Section 3.7.1 in Liermann et al., Introduction in Machine Learning (2019b).

  9. 9.

    H2o.ai is a machine learning framework (see H2O.ai 2019).

  10. 10.

    See Section 4.1 in Liermann et al., Introduction in Machine Learning (2019b).

  11. 11.

    It was beyond the scope of the POC to establish a proper infrastructure to organize the parameter and hyperparameter storage and reading.

  12. 12.

    See Section 3.3 in Liermann et al., Deep Learning—An Introduction (2019a).

  13. 13.

    The opposite approach would be to start with the technology and incorporate the business divisions once the infrastructure is up and running.

  14. 14.

    Errors are discovered earlier in the process.

  15. 15.

    More steps in the reconciliation are automated.

  16. 16.

    A distinction can be made between a minor error (not relevant) and significant deviation.

  17. 17.

    The total impairment value is composed of up to fifteen different key figures. The deviation is usually connected to one key figure that has not been processed properly.

  18. 18.

    Especially in classification models we applied in the POC.

  19. 19.

    Even if sixteen transactions were predicted falsely, one should bear in mind that over 80,000 were properly identified.

  20. 20.

    Overfitting can occur when the historical data is too similar overSeeSeeArea under the curve time and the model can only predict these patterns but is unable to follow changing patterns.

  21. 21.

    In 2020, the project started to place the results and processes in a productive environment.

  22. 22.

    The employees involved came from a physics and mathematical background.

  23. 23.

    Better up and running infrastructure.

Literature

  • Akhgarnush, Eljar, Fabian Bruse, and Ben Hofer. 2021. “New Project Structure.” In The Digital Journey of Banking and Insurance, Volume I—Disruption and DNA, edited by Volker Liermann and Claus Stegmann. New York: Palgrave Macmillan.

    Google Scholar 

  • H2O.ai. 2019. h2o.ai Overview, January 29. Accessed January 29, 2019. http://docs.h2o.ai/h2o/latest-stable/h2o-docs/index.html.

  • Institute for Statistics and Mathematics of WU. 2020. “The Comprehensive R Archive Network.” dplyr: A Grammar of Data Manipulation. Accessed October 15, 2020. https://CRAN.R-project.org/package=dplyr.

  • Liermann, Volker. 2021. “Overview Machine Learning and Deep Learning Frameworks.” In The Digital Journey of Banking and Insurance, Volume III—Data Storage, Processing, and Analysis, edited by Volker Liermann and Claus Stegmann. New York: Palgrave Macmillan.

    Google Scholar 

  • Liermann, Volker, Sangmeng Li, and Norbert Schaudinnus. 2019a. “Deep Learning—An Introduction.” In The impact of Digital Transformation and Fintech on the Finance Professional, edited by Volker Liermann and Claus Stegmann. New York: Palgrave Macmillan.

    Google Scholar 

  • ———. 2019b. “Introduction in Machine Learning.” In The Impact of Digital Transformation and Fintech on the Finance Professional, edited by Volker Liermann and Claus Stegmann. New York: Palgrave Macmillan.

    Google Scholar 

  • Wickham, Hadley, Romain François, Lionel Henry, and Kirill Müller. 2020. “Tidyverse.” dplyr. Accessed October 15, 2020. https://dplyr.tidyverse.org/articles/dplyr.html.

Download references

Author information

Authors and Affiliations

  1. Helaba, Frankfurt, Germany

    Sören Hartung & Manuela Führer

Authors
  1. Sören Hartung
    View author publications

    You can also search for this author in PubMed Google Scholar

  2. Manuela Führer
    View author publications

    You can also search for this author in PubMed Google Scholar

Corresponding author

Correspondence to Manuela Führer .

Editor information

Editors and Affiliations

  1. ifb SE, Grünwald, Germany

    Volker Liermann

  2. ifb Americas, Inc., Charlotte, NC, USA

    Claus Stegmann

Rights and permissions

Reprints and Permissions

Copyright information

© 2021 The Author(s), under exclusive license to Springer Nature Switzerland AG

About this chapter

Verify currency and authenticity via CrossMark

Cite this chapter

Hartung, S., Führer, M. (2021). AI for Impairment Accounting. In: Liermann, V., Stegmann, C. (eds) The Digital Journey of Banking and Insurance, Volume I. Palgrave Macmillan, Cham. https://doi.org/10.1007/978-3-030-78814-8_6

Download citation

  • DOI: https://doi.org/10.1007/978-3-030-78814-8_6

  • Published:

  • Publisher Name: Palgrave Macmillan, Cham

  • Print ISBN: 978-3-030-78813-1

  • Online ISBN: 978-3-030-78814-8

  • eBook Packages: Economics and FinanceEconomics and Finance (R0)