International Journal of Computational Economics and Econometrics (5 papers in press)

Regular Issues

• Modelling and forecasting financial volatility: hybrid econometric-deep learning architectures  
  by Burç Arslan Kaleli, Ahmet Özçam 
  Abstract: Volatility forecasting plays a vital role in financial markets, particularly in asset pricing, portfolio allocation, and derivative valuation. This study proposes hybrid LSTM-based models, namely LSTM-GARCH, LSTM-EGARCH and LSTM-FIGARCH, for forecasting the volatility of the S&P 500 index. Using unidirectional and bidirectional long short-term memory (LSTM) architectures together with gated recurrent units (GRU), model performance is evaluated across different window lengths (7, 30 and 60 days) and forecast horizons (1, 14 and 21 days). The LSTM architecture is designed to balance simplicity and performance while capturing sequential patterns in financial time series. Based on out-of-sample prediction loss, the hybrid models achieve lower forecast errors than traditional approaches across multiple settings, although this improvement may partly reflect the use of a richer input structure. The findings suggest that hybrid deep learning and econometric frameworks offer meaningful improvements in volatility forecasting and can serve as effective tools for researchers and practitioners.
  Keywords: volatility forecasting; GARCH; LSTM; deep learning; hybrid econometric models; financial time series.
  DOI: 10.1504/IJCEE.2026.10080056
   
  
• Adaptive technical analysis as a regime-contingent decision-support model: evidence from an emerging equity market  
  by Mudit Gera 
  Abstract: This study evaluates technical analysis as a regime-contingent decision-support framework rather than as a universal forecasting or abnormal-return generating tool. Using daily NIFTY 50 data from January 2020 to March 2025, it compares five algorithmic decision models: trend-following, oscillator-based, volatility-momentum hybrid, volume-based, and behavioural-anchor rules. Performance is assessed through risk-adjusted metrics, CAPM inference, rolling-window analysis, bootstrap resampling, parameter-sensitivity tests, and regime-specific evaluation. Results show that volume-based and volatility-sensitive models provide episodic economic usefulness during market stress and regime transitions, whereas trend-following rules perform primarily during sustained directional persistence. Behavioural-anchor heuristics, such as Fibonacci retracements, lack statistically robust decision-support value. The findings distinguish conditional economic usefulness from persistent alpha generation and support the adaptive markets hypothesis by showing that technical indicators are best understood as state-contingent decision aids aligned with volatility, liquidity, and market regimes. The study offers practical implications for traders, portfolio managers, and decision-system designers seeking context-aware analytical tools.
  Keywords: technical analysis; regime-contingent modelling; adaptive markets hypothesis; AMH; computational econometrics; decision-support systems; CAPM alpha; emerging equity markets; NIFTY 50; algorithmic trading.
  DOI: 10.1504/IJCEE.2026.10080192
   
  
• Wavelet-fuzzy predictive framework for energy market dynamics: evidence from crude oil data  
  by Prabhat Mittal 
  Abstract: This study develops an adaptive wavelet-fuzzy framework to forecast crude oil price dynamics by integrating wavelet-based multi-resolution analysis with fuzzy rule-based inference. The model decomposes the nonstationary crude oil return series into distinct frequency components, capturing both long-term structural trends and short-term cyclical fluctuations. Fuzzy clustering enables regime identification bullish, bearish, and reversionary phases allowing adaptive transitions across volatile market conditions. Empirical results using daily crude oil data (January 2021-October 2025) show strong predictive performance with an RMSE of 0.0136, an MAE of 0.0102, and a directional accuracy of 79%, outperforming conventional statistical benchmarks. The results of rolling 20-day metrics confirm the models stability and adaptability across market shifts, whereas most mispredictions coincide with major exogenous shocks. The findings demonstrate that integrating wavelet decomposition with fuzzy intelligence enhances both interpretability and forecasting precision, providing a reliable decision-support tool for energy market analyses and strategic policy planning.
  Keywords: fuzzy; wavelet; crude oil; modelling.
  DOI: 10.1504/IJCEE.2026.10080655
   
  
• Mathematical modelling, sensitivity analysis and optimal control in market dominance: exploring stability and policy strategies  
  by Rajib Kumar Dolai, Manotosh Mandal 
  Abstract: This study develops a mathematical model to examine the market dynamics between large and small firms, with particular focus on market dominance. In many markets, large firms usually hold a stronger position, while small firms often struggle to compete. To study this imbalance, two state variables representing the market shares of large and small firms are considered in the model. The system remains uniformly bounded, and the equilibrium points are found to be locally asymptotically stable. Market shares therefore stay feasible and realistic over time. An optimal control mechanism is also introduced, which may represent government intervention or supportive policies. Under the control strategy, the dominance of large firms becomes weaker, while small firms show better growth and stability. Contour plots, sensitivity analysis, and variation of the control parameter are used to understand market behaviour. Parameters g, a and n show positive influence on the feasibility thresholds, whereas l, m and b have negative effects. Pontryagins maximum principle is further applied to determine the optimal conditions of the system. The study considers a simple equation system, and complex nonlinear interactions are not included. Future studies may use sector-specific data and higher-order equations for better understanding of market dynamics.
  Keywords: dominance dynamics; equilibria; Routh-Hurwitz criteria; sensitivity analysis; Pontryagin’s maximum principle; optimal control; government policy.
  DOI: 10.1504/IJCEE.2026.10080776
   
  
• Financial machine learning: a review of mathematical foundations of machine learning, machine unlearning, and adaptive learning strategies  
  by K.K. Saudath, V. Sujatha 
  Abstract: The adoption of machine learning (ML) has already been transforming the FinTech industry. It offers solutions such as credit scoring, loan approvals, fraud detection, algorithmic trading, risk management, etc. Yet, the dynamic nature of financial markets demands consistent adjustment by overriding the obsolete static ML models and adaptive learning approach using state-of-the-art ML techniques. This study synthesises the mathematical frameworks for financial machine learning (FML) covering supervised learning, unsupervised learning, and reinforcement learning. It underscores the importance of unlearning the biased models in response to legal changes (e.g., data privacy legislation) and concept drift. The authors leverage mathematical techniques such as Kullback-Leibler (KL) divergence and regularisation frameworks to formalise the mathematical mechanism of adaptive learning within ML models for FinTech. Finally, the incorporation of strategies such as active learning, online learning, meta-learning (L2L), federated learning, transfer learning, concept drift adaptation, ensemble learning with model updating, reinforcement learning with adaptive policies (RL-AP), self-supervised learning augments adaptability in dynamic financial contexts. This study develops an integrative synthesis of established mathematical frameworks to provide a unified perspective on unlearning and adaptive learning strategies for FML.
  Keywords: FinTech; financial machine learning; FML; machine unlearning; machine relearning; concept drift adaptation.
  DOI: 10.1504/IJCEE.2026.10080783