Scholarlink

This study investigates the efficacy of hybrid time series forecasting models on the Nifty Auto Sectoral Index by analyzing daily closing price data from August 3, 2024, to August 3, 2025. The baseline model employed is ARIMA (0, 1, 0), selected based on the minimum AIC value of 2826.28. Although the standalone ARIMA model produced moderate forecasting accuracy (RMSE: 314.22, MAE: 274.22, MAPE: 1.15%), the results significantly improved when combined with deep learning techniques. Three hybrid models were constructed: ARIMA-LSTM, ARIMA-GRU, and ARIMA-XGBoost. Among these, ARIMA-GRU yielded the best performance, with an RMSE of 171.44, an MAE of 144.69, and an MAPE of 0.61%. ARIMA-STM also demonstrated strong predictive capabilities, while ARIMA-XGBoost returned zero error values, suggesting either model overfitting or evaluation error that warrants further investigation. This comparative analysis highlights the superiority of hybrid models over traditional statistical forecasting. It highlights the potential of combining classical and modern machine learning techniques to achieve enhanced financial time series forecasting accuracy.

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

Hybrid Forecasting Models for the Nifty Auto Index: A Deep Learning-Enhanced ARIMA Approach

  • 会议论文

  • 发表日期：

    2026年

查看全文
  • M. Nagesh,
  • D. Mallikarjuna Reddy,
  • Y. Vijayakumar,
  • Ibne Afzal Munshi Naser

摘要

This study investigates the efficacy of hybrid time series forecasting models on the Nifty Auto Sectoral Index by analyzing daily closing price data from August 3, 2024, to August 3, 2025. The baseline model employed is ARIMA (0, 1, 0), selected based on the minimum AIC value of 2826.28. Although the standalone ARIMA model produced moderate forecasting accuracy (RMSE: 314.22, MAE: 274.22, MAPE: 1.15%), the results significantly improved when combined with deep learning techniques. Three hybrid models were constructed: ARIMA-LSTM, ARIMA-GRU, and ARIMA-XGBoost. Among these, ARIMA-GRU yielded the best performance, with an RMSE of 171.44, an MAE of 144.69, and an MAPE of 0.61%. ARIMA-STM also demonstrated strong predictive capabilities, while ARIMA-XGBoost returned zero error values, suggesting either model overfitting or evaluation error that warrants further investigation. This comparative analysis highlights the superiority of hybrid models over traditional statistical forecasting. It highlights the potential of combining classical and modern machine learning techniques to achieve enhanced financial time series forecasting accuracy.