Aplication of Artificial Neural Networks for Forecasting in Business Economy
Aplication of Artificial Neural Networks for Forecasting in Business Economy

Author(s): Andrea Kolková
Subject(s): Social Sciences, Economy
Published by: Vysoká škola ekonomická v Praze
Keywords: Artificial neural networks; exponential smoothing; ARIMA; BATS
Summary/Abstract: Purpose: The forecasting method is now a whole lot. They are often based on the specificconditions of the given time series, and their methodology is mostly the result of research inscientific centres and universities. In recent years, Artificial Intelligence has been very muchdiscussed (hereafter AI). Implementation of AI into enterprise decision-making brings a wholehost of new opportunities and challenges. One of them is certainly the use of AI in forecasting. Design/methodology/approach: The paper after classic models present, the AI-based model,namely the neural networks model, introduce. Subsequently, the models are applied to 166monthly data from year 2008 to 2018. After analysing the data, forecasting ex-post is performedand evaluated according to selected accuracy indicators. After evaluating accuracy, the mostaccurate model for the given enterprise variables is selected and ex-ante forecasting performed. Findings: Benefit of this paper can be seen in particular in the expansion of possible forecastingmethods to ensure the most accurate results of business forecasts. The evaluation of thesuitability of the models is ensured by the best values of the selected accuracy measures. Research/practical implications: The paper confirms the possibilities of using the neuralnetwork method for business time series as the best model with RMSE 0.3134768. In thepractice of specific businesses, the contribution can help with the selection of suitable methodsfor forecasting. In future research, I can focus on other forecasting methods, such as the use ofother AI tools, chaos theory, fuzzy logic, or genetic algorithms. Originality/value: At present, the practical use of neural networks in the corporate economyin the Czech Republic is still an outlying issue. Its wider use in practice requires exploration ofthe use of academic and other scientific institutions. The way in which scientific knowledgecan be accessed through practice can be a wider use of this tool in practice.

• Details
• Contents