Page 109 - Special Topic Session (STS) - Volume 4

STS566 K. Prokopenko et al.
2
against several reliability output parameters. Especially the R shows that the
new CSARMA model is more stable and accurate when applied for this data.
It is foreseen that with more historical data the ML algorithms will also detect
seasonality better. Since CSARMA was designed based on research and strictly
tuned on the specific data structure it is most likely that CSARMA will be less
reliable than ML if for some reason seasonality changes. Therefore
Giesecke+Devrient performs complete data research and produces
forecasting algorithms as part of the software suite that are optimal for a given
data structure. Central and Commercial banks can benefit from the high-
quality forecasts for improving their expected cash flows based on end
customer behaviour. Giesecke+Devrient will further extend the forecasting
capabilities with inputs about the fitness levels of cash using data from
counting machines inside cash centres.

References
1. Fisher RA. Statistical Methods and Scientific Inference. Ed 2 (rev)
Edinburgh, UK: Oliver and Boyd; 1959
2. G E P Box, G M Jenkins and G C Reinsel 2000 Time Series Analysis -
Forecasting and Control Prentice Hall New Jersey 1994
3. Sepp Hochreiter; Jürgen Schmidhuber (1997). "Long short-term memory".
Neural Computation. 9 (8): 1735–1780. doi:10.1162/neco.1997.9.8.1735.
PMID 9377276.
4. Chen, Tianqi; Guestrin, Carlos (2016). "XGBoost: A Scalable Tree Boosting
System". In Krishnapuram, Balaji; Shah, Mohak; Smola, Alexander J.;
Aggarwal, Charu C.; Shen, Dou; Rastogi, Rajeev (eds.). Proceedings of the
22nd ACM SIGKDD International Conference on Knowledge Discovery
and Data Mining, San Francisco, CA, USA, August 13-17, 2016. ACM. pp.
785–794.
5. https://www.nbg.gov.ge/uploads/publications/annualreport/2018/eng_a
nnual_2017.pdf (June 25, 2019)
6. https://www.npr.org/2019/02/11/691334123/swedens-cashless-
experiment-is-it-too-much-too-fast?t=1559903488540 (June 7, 2019)

98 | I S I W S C 2 0 1 9