Open research questions in Forecasting Techniques and Applications

What the literature leaves open

• FUTURE WORK (RECOMMENDATIONS) Future research should investigate automated contract-testing frameworks capable of detecting upstream schema evolution before ETL execution.

  A Modular Data Integration Architecture for Multi-Driver Economic Forecasting Models · 2026 · DOI

• Future research should focus on expanding the dataset with higher-frequency or real-time IoT-based observations and integrating advanced AI models such as Artificial Neural Networks (ANNs) to capture nonlinear dependencies and improve predictive performance.

  A framework for carbon footprint computation and forecasting for Nigeria’s industrial decarbonization plan (NIDP) · 2026 · DOI

• The future research will further optimize and improve the VMD-PSO-BiLSTM regression prediction algorithm proposed in this paper, expand its application scenarios and practical value, and make up for the shortcomings of existing research. First of all, at the level of feature construction, multi-source external correlation features will be introduced, including e-commerce platform promotion activities, market consumption trends, macroeconomic data, logistics and transportation efficiency and emergencies, etc., to build a more comprehensive feature input system, solve the problem that the current model can't capture external disturbance factors enough, and further enhance the model's adaptability to nonlinear and non-stationary supply chain demand fluctuations. Secondly, in the aspect of model optimization, the attention mechanism and lightweight network structure will be combined to optimize the parameter adaptive selection of VMD decomposition, and at the same time, the optimization strategy of PSO algorithm will be improved, so as to reduce the model training time and calculation cost, improve the model reasoning speed, and meet the actual needs of real-time forecasting and engineering deployment of supply chain. In addition, the model application scenarios will be expanded, from single-category demand forecasting to collaborative forecasting of multi-category and multi-supply chain nodes, and combined with federated learning technology, it will effectively solve pain points such as data islands and privacy leakage of upstream and downstream enterprises in the supply chain, and realize data security sharing and joint modeling. Finally, the robustness of the model under sudden demand disturbance and extreme 241 Proceedings of the 4th International Conference on Mathematical Physics and Computational Simulation DOI: 10.54254/2753-8818/2026.34453 market environment will be studied, and the uncertainty quantification method will be introduced to improve the interpretability of the model. At the same time, it will be verified in e-commerce supply chain scenarios in different regions and scales, so as to further improve the generalization ability of the model and provide more stable, reliable and comprehensive intelligent decision-making support for refined and intelligent management of supply chain.

  Precise Demand Forecasting for Supply Chain Based on Machine Learning Algorithms · 2026 · DOI

• This paper presents a phase- and data-aware metric table that situates machine learning within established estimate classes for LCC estimation in PSS. Across the reviewed evidence, model choice aligns with project maturity and feature availability. Transparent parametric baselines remain useful when inputs are sparse, while simple data-driven models become effective once a compact, validated feature set exists. Richer non-linear approaches gain advantage as data availability, model governance, and process maturity increase. Importantly, the findings indicate that a combination of traditional and machine learning methods can be beneficial, particularly when leveraging complementary strengths such as transparency, robustness, and predictive flexibility. However, it remains an open research task to systematically evaluate which method combinations yield meaningful synergies, under which life cycle phases, data conditions, and decision contexts. Hybrid setups already show added value for in-service planning, whereas end-of-life estimation remains underexplored. The proposed metric table serves as a deployment guide that indicates when and how ML methods can be applied, which trade-offs to expect, and where hybrid approaches may be most promising. Further progress is likely to arise from more consistent metric definitions and validation protocols, shared concept-phase datasets with agreed feature taxonomies, and increased attention to interpretability and human-in-the-loop integration alongside accuracy. Extending analyses to neglected life cycle stages and stress-testing models across domains will be essential to assess generalizability. In practice, the metric table supports more transparent planning and decision-making by aligning method selection with project definition, data readiness, and intended use, while indicating potential opportunities for combining traditional and machine learning approaches that require further empirical evaluation.

  Life cycle cost estimation in product-service systems: a review of machine learning methods · 2026 · DOI

• The present study is limited by the size of the per-scope datasets, which restricts the SVR model's ability to fully exploit its non-linear capacity, and by its reliance on a single e-commerce operation's data.

  PHARMACEUTICAL E-COMMERCE SALES FORECASTING USING GM (1,1) AND SVR · 2026 · DOI

• The study is limited to a single pharmaceutical e-commerce dataset with monthly observations. Future work will compare the proposed approach with LSTM, Prophet, XGBoost and Transformer-based forecasting models using larger multi-store datasets and additional external factors such as promotions and holidays. VI. REFERENCES Hong T. Pharmaceutical E-Commerce Sales Forecasting Using GM(1,1) and SVR. 2025 2nd International Conference on Intelligent Computing and Robotics (ICICR), 2025, pp. 787–794. doi: 10.1109/ICICR65456.2025.00140. Jiang J, Zhang M, Huang Z. A new adaptive grey prediction model and its application. Alexandria Engineering Journal, 2025, 120: 515–522. doi: 10.1016/j.aej.2025.02.027. Tien T L. A research on the grey prediction model GM(1,n). Applied Mathematics and Computation, 2012, 218(9): 4903–4916. doi: 10.1016/j.amc.2011.10.055. Shi P, Xu L, Qu S, et al. Assessment of hybrid kernel function in extreme support vector regression model for streamflow time series forecasting based on a Bayesian estimator decomposition algorithm. Engineering Applications of Artificial Intelligence, 2025, 149: 110514. doi: 10.1016/j.engappai.2025.110514. Mahin M P R, Shahriar M, Das R R, et al. Enhancing Sustainable Supply Chain Forecasting Using Machine Learning for Sales Prediction. Procedia Computer Science, 2025, 252: 470–479. doi: 10.1016/j.procs.2025.01.006. Ahamed S F, Vijayasankar A, Thenmozhi M, et al. Machine learning models for forecasting and estimation of business operations. The Journal of High Technology Management Research, 2023, 34(1): 100455. doi: 10.1016/j.hitech.2023.100455.

  PHARMACEUTICAL E-COMMERCE SALES FORECASTING USING GM (1,1) AND SVR · 2026 · DOI

• Several limitations of this study should be acknowledged. First, the analysis relies on annual national-level data covering the period 2004–2022. While this long-term perspective is suitable for strategic budget planning, the use of annual aggregates may mask short-term fluctuations and regional heterogeneity in maintenance and repair costs. Future research could address this limitation 103 THE BALTIC JOURNAL OF ROAD AND BRIDGE ENGINEERING2026/21(1)Haydar Gundogdu, Omer Faruk Cansiz, Mehmet Fatih CanPrediction of Road Maintenance and Repair Costs in Turkey Using a Hybrid Modelling Approach by employing higher-frequency data (e.g., monthly or quarterly) and regionally disaggregated datasets to capture spatial and temporal variations more explicitly. Second, although the study integrates multiple regularisations, variable- selection, and hybrid modelling techniques to mitigate overfitting risks, the relatively limited number of observations inherent in annual time-series data may still constrain model generalisability. Expanding the dataset with additional temporal coverage or complementary cross-sectional information could further strengthen the robustness of future analyses. Third, potential structural breaks arising from extraordinary events – such as economic crises, policy reforms, or the COVID-19 pandemic – are not modelled explicitly in the current framework. While hybrid and non-linear methods partially accommodate such dynamics, future studies could incorporate regime-switching models or structural break tests to better account for abrupt changes in cost behaviour. Finally, the analysis focuses primarily on financial, infrastructural, economic, and meteorological determinants of MRC. Institutional, managerial, and contractual factors – such as procurement practices, maintenance scheduling efficiency, and contractor performance – are not explicitly considered due to data limitations. Incorporating such variables in future research may provide a more comprehensive understanding of cost dynamics and further enhance the policy relevance of the modelling framework.

  Prediction of Road Maintenance and Repair Costs in Turkey Using a Hybrid Modelling Approach · 2026 · DOI

• We propose \method{}, a sparse mixture-of-experts framework that represents each series with a multidimensional forecastability fingerprint, mines expert-suitability targets from validation performance, and trains a cost-aware sparse router to activate a small budgeted set of experts for each series.

  FAME: Forecastability-Aware Mixture of Experts for Heterogeneous Time Series Forecasting · 2026

• This review is addressed to a broad readership: academic researchers and early career scholars tracing the intellectual and methodological history of the field, practitioners seeking reliable and transparent methods for operational forecasting, and management mathematicians and quantitative analysts from cognate disciplines—operational research, decision analytics, and management science—for whom benchmark forecasting methods constitute an essential but underexplored toolkit.

  Benchmarks in univariate time series forecasting: a historical and methodological review · 2026 · DOI

• Future work will extend SHOS toward adaptive parameter learning, automated lifecycle-aware representations, and integration with prescriptive inventory-optimization frameworks. Further evaluation across multi-echelon supply chains, real-time deployment environments, and additional industry sectors will help establish broader generalizability and operational impact. The present study evaluates SHOS representations under a fixed monthly aggregation framework. Although monthly discretization improves statistical stability under highly sparse demand conditions, alternative temporal resolutions may preserve additional renewal dynamics and short- term demand structure. Future work should therefore investigate weekly ARTICLE IN PRESS ARTICLE IN PRESS ACCEPTED MANUSCRIPT aggregation, adaptive temporal discretization, and multi-resolution intermittent-demand representation frameworks to evaluate the sensitivity of SHOS-derived features to aggregation bucket size across different industrial forecasting environments. The present evaluation framework focuses primarily on aggregate forecasting accuracy and representation-oriented comparative analysis under sparse-demand conditions. Future work should extend the evaluation protocol to include event-oriented intermittent-demand metrics, such as the sensitivity and specificity of nonzero-demand detection, conditional evaluation restricted to positive-demand observations, and Average Detection Delay (ADD) to measure responsiveness to demand reactivation after prolonged inactivity periods. Future work should also investigate alternative representation-learning approaches for sparse dealer-part interaction matrices, including zero- inflated probabilistic matrix factorization, Poisson factorization, and nonnegative matrix factorization methods specifically designed for highly sparse count-based environments. Such approaches may further improve latent behavioural representation under extreme intermittent-demand conditions.

  Event-preserving feature engineering for intermittent demand forecasting using SHOS · 2026 · DOI

• Although the proposed Transformer-LSTM and Soft Actor- Critic based dynamic pricing framework demonstrates strong performance, several directions can be explored to further enhance its capabilities. Page 486 www.rsisinternational.org INTERNATIONAL JOURNAL OF LATEST TECHNOLOGY IN ENGINEERING, MANAGEMENT & APPLIED SCIENCE (IJLTEMAS) ISSN 2278-2540 | DOI: 10.51583/IJLTEMAS | Volume XV, Issue V, May 2026 One important extension is the incorporation of multi-agent reinforcement learning, where multiple competing sellers dy- namically adjust prices in a shared market environment. This would enable the system to model realworld competitive pricing scenarios more effectively. Another promising direction is the integration of uncertainty-aware forecasting techniques, such as Bayesian deep learning or probabilistic Transformers, to better quantify prediction confidence and improve robustness under highly volatile demand conditions. The current system assumes a single-product or independent pricing setup. Future work can extend the framework to multi-product pricing with cross-elasticity modeling, where the demand of one product depends on the pricing of related products. In addition, fairness-aware pricing and ethical constraints can be incorporated to ensure that pricing strategies remain transparent and do not lead to unintended price discrimination or regulatory concerns. From a system perspective, deploying the framework in a real-world production environment with live user traffic would provide valuable insights into performance under real- time constraints. Integration with edge computing or streaming pipelines could further reduce latency and improve scalability. Finally, advanced techniques such as causal inference and offline reinforcement learning can be explored to improve sam- ple efficiency and enable learning from limited or historical data without requiring extensive online interaction. These directions provide opportunities to further improve the adaptability, robustness, and real-world applicability of dynamic pricing systems. REFERENCES 1) J. Liu et al., “Dynamic Pricing on E-Commerce Platform with Deep Reinforcement Learning: A Field Experiment,” arXiv preprint arXiv:1912.02572, 2021. 2) H. Yin and Q. Han, “Dynamic Pricing Model of E-Commerce Plat- forms Based on Deep Reinforcement Learning,” Computer Modeling in Engineering & Sciences, 2021. 3) J. Sun et al., “Dynamic Pricing Model for E-Commerce Products Based on DDQN,” 2024. A. Holovko and T. Firman, “Batch Reinforcement Learning for Dynamic Pricing,” 2021. 4) F. Lange et al., “Reinforcement Learning vs Dynamic Programming for Pricing,” 2025. 5) L. Guo and X. Zhang, “Dynamic Pricing using LSTM,” IEEE Access, 2025. 6) S. Kumar et al., “Weight Optimized LSTM for Pricing,” 2023. 7) L.

  Dynamic Price Allocation and Optimization for E-Commerce Platforms Using Reinforcement Learning and Deep Learning · 2026 · DOI

• Technical sectors should replace subjective model choices with multicriteria analysis methods such as CRITIC to ensure statistically robust selection based on variability and conflict of multiple metrics.

  Combining demand classification and forecasting models in spare parts inventory management for the energy sector · 2026 · DOI

• The assumption that 50 percent of cotton imports are from developed countries and 50 percent from developing countries is a simplification that may not reflect actual trade patterns.

  Empirical and Machine Learning Forecasting for Offline Retail: Nonlinear Weather Effects and Heterogeneity · 2026 · DOI

• The results of this study imply a need for further studies on the applications of TBATS models in forecasting the required cash level in ATMs, which in turn may help improve the efficiency of ATMs network management.

  Using trigonometric seasonal models in forecasting the size of withdrawals from automated teller machines · 2023 · DOI

• Despite the fact that construction costs can vary widely across locations with different market conditions and environments, the national CCI, the simple average of construction costs for 20 US metropolitan areas, has often been used to forecast CCIs across the nation.

  Predicting City-Level Construction Cost Index Using Linear Forecasting Models · 2020 · DOI

• The purpose of this additional study on automobile demand is (1) to point out the forecasting accuracy of the market segmentation approach and (2) to estimate the effects of the energy crisis on automobile demand for the years 1979-83.

  Statistical Demand Functions for Automobiles and Their Use for Forecasting in an Energy Crisis · 1980 · DOI

• What remains to be done is, above all, to make forecasting an instrument for the contemplation and preparation of profound changes--not only in the structure of social systems, but also in the ethics of which this structure is an expression, the "ethics of whole systems" (Churchman);6s and beyond that, in the general cultural basis.

  Forecasting and the systems approach: A critical survey · 1972 · DOI

• The study contributes to the under-researched domain of inventory management in the Indian home-lift and elevator implementable recommendations for reducing inventory carrying costs, improving warehouse utilization, and strengthening service reliability.

  Inventory Optimization Using FSN Analysis in Spare Parts Planning · 2026 · DOI

• Sectoral practices must evolve from the use of small samples to the processing of the total population of data (Big Data) to eliminate selection biases and increase strategic reliability.

  Combining demand classification and forecasting models in spare parts inventory management for the energy sector · 2026 · DOI

• It is recommended to constantly reevaluate the models whenever new scenarios or data emerge, since no model is ideal for all types of historical series permanently.

  Combining demand classification and forecasting models in spare parts inventory management for the energy sector · 2026 · DOI

• Import from India is not considered in the scenario analysis, despite India's expanding textile industry, limiting the comprehensiveness of the global cotton trade assessment.

  Empirical and Machine Learning Forecasting for Offline Retail: Nonlinear Weather Effects and Heterogeneity · 2026 · DOI

• Environmental impacts caused by defoliant and plastic film are neglected because cotton planting and picking methods vary among countries and there is no common basis for comparison.

  Empirical and Machine Learning Forecasting for Offline Retail: Nonlinear Weather Effects and Heterogeneity · 2026 · DOI

• Due to a lack of U.S. data about the rank and ratio of agrochemical toxicity, the number is assumed to be similar to that in another developed country: Australia.

  Empirical and Machine Learning Forecasting for Offline Retail: Nonlinear Weather Effects and Heterogeneity · 2026 · DOI

• One of the major issues being faced in demand forecasting is insufficient forecast accuracy to predict the expected demand and fluctuation in actual vs.

  Prediction of Intermittent Demand Occurrence using Machine Learning · 2024 · DOI

• A typical retail setting involves predicting the demand for hundreds of items simultaneously, some with abundant historical data and others with scarce data.

  Data Aggregation and Demand Prediction · 2022 · DOI

Most-cited papers in Forecasting Techniques and Applications

• Recurrent Neural Networks for Time Series Forecasting: Current status and future directions · International Journal of Forecasting · 2020 · 1,115 citations
• Forecasting: theory and practice · International Journal of Forecasting · 2022 · 726 citations
• The Effect of Collaborative Forecasting on Supply Chain Performance · Management Science · 2001 · 414 citations
• M5 accuracy competition: Results, findings, and conclusions · International Journal of Forecasting · 2022 · 388 citations
• A machine learning based approach for predicting blockchain adoption in supply Chain · Technological Forecasting and Social Change · 2020 · 259 citations
• Forecasting wholesale prices of yellow corn through the Gaussian process regression · Neural Computing and Applications · 2024 · 258 citations
• Forecast combinations: An over 50-year review · International Journal of Forecasting · 2022 · 233 citations
• Kaggle forecasting competitions: An overlooked learning opportunity · International Journal of Forecasting · 2020 · 222 citations
• Is Mamba effective for time series forecasting? · Neurocomputing · 2024 · 214 citations
• Peanut oil price change forecasts through the neural network · foresight · 2025 · 189 citations

Most recent work

• The AI Energy-Demand Forecast Scorecard: a reproducible audit of how the field forecasts data-centre electricity demand · Zenodo (CERN European Organization for Nuclear Research) · 2026
• CLASNet: A Cognitive Load–Aware CNN-LSTM-Attention Framework for Supply Chain Demand Forecasting and Adaptive Human–Computer Interaction · Journal of Computer Science and Frontier Technologies · 2026
• Empirical and Machine Learning Forecasting for Offline Retail: Nonlinear Weather Effects and Heterogeneity · Journal of Systems Science and Systems Engineering · 2026
• Combining demand classification and forecasting models in spare parts inventory management for the energy sector · The International Journal of Advanced Manufacturing Technology · 2026
• Simulation Studies on the Impact of Using Demand Forecasts Based on Deep Neural Networks for Inventory Replenishment Efficiency · Logforum · 2026
• Demand Forecasting Strategies for New Product Launches in the Consumer Electronics Sector: A 2025 Perspective · Zenodo (CERN European Organization for Nuclear Research) · 2026
• An autoregressive lag aware sparse transformer framework for demand forecasting · Discover Artificial Intelligence · 2026
• Benchmarks in univariate time series forecasting: a historical and methodological review · IMA Journal of Management Mathematics · 2026
• An AI-Based Approach to Food Demand Forecasting Incorporating External Market Factors · International Journal of Advanced Research in Science Communication and Technology · 2026
• Sales Forecast Prediction Using Machine Learning · International Journal of Innovative Research in Engineering · 2026

Related topics in Decision Sciences