EUVIMEDEUVIMED entdecken
Crossref

Investigating the Determinants of Stock Prices and Comparing the Performance of Advanced Nonlinear Methods in Forecasting

Parasto Azami, Najmeh Kargar Kamvar, Hoda Hemmati

Dynamic Management in Business Analysis · 2027 · 6 (2) · 1-19

DOI: 10.61838/dmbaj.374 ↗

Kurz gefasst

Fachartikel in der Zeitschrift Dynamic Management in Business Analysis, 2027. Verfasst von Parasto Azami und 2 weiteren Autor:innen.

AbstractBibliografische AngabenZitierenVerwandte Studien

Abstract (Originalsprache)

Objective

The objective of this study was to compare the predictive accuracy and performance of fuzzy regression, genetic algorithm, support vector machine, decision tree, and machine learning methods in forecasting stock closing prices using real out-of-sample observations. Methodology: This applied quantitative study utilized daily stock market data of Isfahan Steel Company from January 1, 2024, to November 19, 2025. After removing highly collinear variables, selected predictors were incorporated into the forecasting models. The final 30 observations were reserved as real out-of-sample data to evaluate forecasting performance. Fuzzy regression, genetic algorithm, interpretable machine learning, decision tree, and support vector machine models were developed and compared. Prediction accuracy was assessed using Mean Squared Error (MSE), Root Mean Squared Error (RMSE), and Mean Absolute Error (MAE).

Findings

The results demonstrated significant differences in predictive performance among the examined models. Fuzzy regression achieved the highest forecasting accuracy, producing the lowest error values (MSE = 26, RMSE = 31, MAE = 996). Interpretable machine learning ranked second (MSE = 32, RMSE = 40, MAE = 1,567). Decision tree models showed moderate performance (MSE = 53, RMSE = 71, MAE = 5,027). In contrast, support vector machines and genetic algorithms generated substantially larger forecasting errors and exhibited weaker generalization ability on out-of-sample data. Furthermore, opening price, highest price, and lowest price emerged as the most influential predictors across most forecasting models.

Conclusion

The findings indicate that interpretable forecasting approaches, particularly fuzzy regression, provide a superior balance between predictive accuracy and economic interpretability. These models not only deliver reliable forecasts but also facilitate a clearer understanding of the relationships among financial variables. Consequently, they can serve as effective tools for short-term stock price forecasting and evidence-based investment decision-making in capital markets.

Zitieren

APA

Azami, P., Kamvar, N. K., & Hemmati, H. (2027). Investigating the Determinants of Stock Prices and Comparing the Performance of Advanced Nonlinear Methods in Forecasting. Dynamic Management in Business Analysis, 6(2), 1-19. https://doi.org/10.61838/dmbaj.374

Vancouver

Azami P, Kamvar NK, Hemmati H. Investigating the Determinants of Stock Prices and Comparing the Performance of Advanced Nonlinear Methods in Forecasting. Dynamic Management in Business Analysis. 2027;6(2):1-19. doi:10.61838/dmbaj.374

BibTeX
@article{Azami2027,
  author = {Azami, Parasto and Kamvar, Najmeh Kargar and Hemmati, Hoda},
  title = {Investigating the Determinants of Stock Prices and Comparing the Performance of Advanced Nonlinear Methods in Forecasting},
  journal = {Dynamic Management in Business Analysis},
  year = {2027},
  volume = {6},
  number = {2},
  pages = {1--19},
  doi = {10.61838/dmbaj.374}
}
Kostenlos starten →