Financial time series forecasting using twin support vector regression.

Area of Science:

• * Computational Finance
• * Machine Learning
• * Econometrics

Background:

• * Financial time series forecasting is vital for robust decision-making.
• * Noisy data and non-stationarity pose significant challenges in prediction.
• * Existing methods may struggle with these inherent data complexities.

Purpose of the Study:

• * To propose an effective method for financial time series prediction.
• * To address the challenges of noisy data and non-stationarity.
• * To evaluate the performance of the proposed method across diverse financial datasets.

Main Methods:

• * Implementation of twin support vector regression (TSVR).
• * Application of TSVR to various financial time series datasets (IT, stock market, banking, oil/petroleum).
• * Performance evaluation using root mean squared error and standard deviation.

Main Results:

• * Twin support vector regression effectively handles noisy and non-stationary financial data.
• * Numerical experiments across multiple industries validate the method's applicability.
• * The proposed TSVR method shows improved prediction accuracy compared to standard methods.

Conclusions:

• * Twin support vector regression is a viable and effective tool for financial time series forecasting.
• * The method offers computational advantages, being faster than standard support vector regression.
• * This approach enhances the robustness of financial decision-making through improved prediction accuracy.