Scholarlink

<p>Financial forecasting on stock market returns has always posed challenges for investors, economists, and policymakers. Various forecasting modelling techniques are employed to obtain accurate data predictions, but most provide limited insights. This study systematically evaluates the Spline-GARCH model with the GARCH family, Neural Network (NN), and Machine Learning (ML) techniques. Specifically, Random Forest (RF), Support Vector Machine (SVM), Nonlinear Autoregressive Neural Networks (NARNN), and Long-Short Term Model (LSTM) models are employed to predict stock market returns. The recursive forecasting method is used to forecast daily returns of nine selected stock market indices of the Belt and Road Initiative linked with the China-Pakistan Economic Corridor (CPEC) from January 2015 to September 2024. The models’ forecasting performance is assessed using various forecast error measurement criteria. The empirical results revealed that the Spline-GARCH is the best-fitted model compared to other GARCH family models in all stock markets, with a high significance level in seven forecast evaluation criteria. The forecasting performance of the Spline-GARCH model outperforms that of all NN and ML models in the stock markets of Kazakhstan, Malaysia, Pakistan, Russia, and Saudi Arabia. This paper explicitly demonstrated that the Spline-GARCH model is a better ML model as a superior alternative for accomplishing accurate financial forecasts of stock market returns of BRI countries. To improve the accuracy of financial forecasts and guide sound investment decisions, particularly within Belt and Road Initiative (BRI) countries, economists are advised to incorporate spline-GARCH modelling into their analytical toolkit.</p>

错误：搜索内容不能为空，请输入英文关键词
错误：关键词超出字数限制，请精简
高级检索

Financial forecasting and new frontiers of Spline-GARCH: a superiority analysis over the traditional GARCH and machine learning models on belt and road initiative economies

  • 期刊论文

  • 发表日期：

    2026年5月4日

查看全文
  • Tayyab Raza Fraz,
  • Samreen Fatima,
  • Fayaz Hussain Tunio,
  • Magdalena Radulescu

摘要

Financial forecasting on stock market returns has always posed challenges for investors, economists, and policymakers. Various forecasting modelling techniques are employed to obtain accurate data predictions, but most provide limited insights. This study systematically evaluates the Spline-GARCH model with the GARCH family, Neural Network (NN), and Machine Learning (ML) techniques. Specifically, Random Forest (RF), Support Vector Machine (SVM), Nonlinear Autoregressive Neural Networks (NARNN), and Long-Short Term Model (LSTM) models are employed to predict stock market returns. The recursive forecasting method is used to forecast daily returns of nine selected stock market indices of the Belt and Road Initiative linked with the China-Pakistan Economic Corridor (CPEC) from January 2015 to September 2024. The models’ forecasting performance is assessed using various forecast error measurement criteria. The empirical results revealed that the Spline-GARCH is the best-fitted model compared to other GARCH family models in all stock markets, with a high significance level in seven forecast evaluation criteria. The forecasting performance of the Spline-GARCH model outperforms that of all NN and ML models in the stock markets of Kazakhstan, Malaysia, Pakistan, Russia, and Saudi Arabia. This paper explicitly demonstrated that the Spline-GARCH model is a better ML model as a superior alternative for accomplishing accurate financial forecasts of stock market returns of BRI countries. To improve the accuracy of financial forecasts and guide sound investment decisions, particularly within Belt and Road Initiative (BRI) countries, economists are advised to incorporate spline-GARCH modelling into their analytical toolkit.