Integrating Statistical and Neural Forecasting Paradigms: Comparative Predictive Efficacy of Theta Modelling and Neural Network–Enhanced ARIMA for Indian Retail - Gold Prices - Evidence from daily retail gold price data in India, 2014–2025

Rajib Bhattacharya

doi:10.62127/aijmr.2025.v03i05.1146

Integrating Statistical and Neural Forecasting Paradigms: Comparative Predictive Efficacy of Theta Modelling and Neural Network–Enhanced ARIMA for Indian Retail - Gold Prices - Evidence from daily retail gold price data in India, 2014–2025

Author(s)	Rajib Bhattacharya
Country	India
Abstract	This study presents a comparative analysis of two distinct forecasting paradigms—the Theta model and the Neural Network–Enhanced ARIMA (ARIMA–NN) model—to evaluate their predictive efficiency in forecasting daily retail gold prices in India over the period 2014–2025. Gold, a crucial financial and cultural asset in India, exhibits nonlinear, volatile, and cyclical dynamics driven by macroeconomic factors, investor sentiment, and seasonal demand. Traditional statistical approaches like ARIMA and Theta modelling effectively capture trend and curvature but often fail to represent nonlinearities and structural breaks. This study, therefore, examines whether hybridizing econometric rigour with neural computation enhances short-term forecasting precision and adaptability in volatile markets. The methodology involves applying the Theta model—a decomposition-based statistical technique that extracts and combines multiple curvature components—and a hybrid ARIMA–NN model, wherein the residuals from an ARIMA forecast are modelled using a feedforward neural network. Both frameworks are evaluated using Mean Absolute Error (MAE), Root Mean Square Error (RMSE), and Mean Absolute Percentage Error (MAPE), supplemented by the Diebold–Mariano test for statistical comparison. Empirical results reveal that both models achieve high predictive accuracy, with MAPE values below 5%, confirming their reliability for short-horizon gold price forecasting. However, the ARIMA–NN hybrid consistently outperforms the Theta model, yielding a lower MAPE (4.55% vs 4.86%). The neural augmentation enables the ARIMA–NN model to capture residual nonlinear dependencies, improving responsiveness to market fluctuations and structural irregularities. Conversely, the Theta model exhibits strong trend smoothing but underestimates during rapid price surges due to its deterministic framework. The findings underscore that integrating neural learning into statistical frameworks enhances both accuracy and adaptability, offering a balanced synthesis of interpretability and computational flexibility. The study contributes to the growing body of literature on hybrid forecasting models, demonstrating their potential to advance predictive analytics for volatile commodities like gold, particularly in emerging markets characterized by cyclical and behavioural complexities.
Keywords	Gold Price Forecasting, ARIMA–NN Hybrid Model, Theta Model, Neural Networks, Time-Series Modelling JEL Classification: C22, C45, C53, G17, Q02
Discipline	Other
Published In	Volume 3, Issue 5, September-October 2025
Published On	2025-10-31
Cite This	Integrating Statistical and Neural Forecasting Paradigms: Comparative Predictive Efficacy of Theta Modelling and Neural Network–Enhanced ARIMA for Indian Retail - Gold Prices - Evidence from daily retail gold price data in India, 2014–2025 - Rajib Bhattacharya - AIJMR Volume 3, Issue 5, September-October 2025. DOI 10.62127/aijmr.2025.v03i05.1146
DOI	https://doi.org/10.62127/aijmr.2025.v03i05.1146
Short DOI	https://doi.org/g98ndv

View / Download Full-text Research PDF File

About AIJMR Fees & Payment Current Issue Publication Archive	Submit Research Paper Track Submission Status Publication Guidelines Publication Ethics Peer Review & Plagiarism	Join as a Reviewer Editors & Reviewers Reviewer Referral Program Get Reviewer Membership Certi.	Website/Journal Policies Usage Policy Content Policies Privacy Policy

Contact Us		+91-9328-605090	editor@aijmr.com

Advanced International Journal of Multidisciplinary Research

An Open Access Peer Reviewed Multidisciplinary Bi-monthly Scholarly International Journal

Integrating Statistical and Neural Forecasting Paradigms: Comparative Predictive Efficacy of Theta Modelling and Neural Network–Enhanced ARIMA for Indian Retail - Gold Prices - Evidence from daily retail gold price data in India, 2014–2025

Share this