A Machine Learning Approach to Financial Forecasting: A Case Study

Year 2023, Volume: 32, 8 - 12, 30.11.2023

Ahmet Kara

https://doi.org/10.55549/epess.1412722

Abstract

This paper undertakes a machine learning-based forecasting of a subset of financial processes pertaining to the stock market for a particular period in Turkey. There are various machine learning/artificial intelligence algorithms ranging from multilayer perceptron to support vector machines that can be used, with varying degrees of success, for forecasting purposes. The forecasting task to be undertaken in this paper will be carried out in contexts inclusive of a number of crisis-associated complexities generating unusual fluctuations in the financial markets. These fluctuations could pose, for traditional methods, significant difficulties that could be predictably overcome by machine learning/artificial intelligence algorithms which could escape a reasonable range of the possible complications that could be encountered. We will employ a number of algorithms which we will compare and contrast in accordance with a chosen performance metric. Not all algorithms perform equally well but some yield results that could be comfortably and successfully used for further analysis. Successful policy analyses addressing some of the essential intricacies of financial processes are of both theoretical and practical significance. They could produce considerable welfare improvements in emerging economies such as Turkey. Possible ways in which such improvements could be modeled are worthy of future research.

Keywords

Stock market, Machine learning, Forecasting

References

• Beker, V.A. (2014), Why should economics give chaos theory another chance? In M. Faggini & A. Parziale (Eds.), Complexity in Economics: Cutting Edge Research (pp. 205-223). Springer: New York.
• Fedyk, T. (2017). Refining financial analysts' forecasts by predicting earnings forecast errors. International Journal of Accounting and Information Management, 25(2), 256-272.
• Ferrara, L., & Guegan, (2000). D. Forecasting financial times series with generalized long memory processes. In C.L. Dunis (Ed.), Advances in Quantitative Asset Management (pp.319-342). Springer: London.