Subtopic Deep Dive
Forecast Combination Techniques
Research Guide
What is Forecast Combination Techniques?
Forecast combination techniques aggregate predictions from multiple forecasting models to produce a single, more accurate forecast.
Methods include simple averaging, variance-based weighting, and stacking. Newbold and Granger (1974) demonstrated that combining forecasts outperforms individual models across economic time series (834 citations). Approximately 10 key papers from 1974-2020 analyze performance under uncertainty.
Why It Matters
Forecast combinations improve accuracy in supply chain planning and pandemic response, as shown by Nikolopoulos et al. (2020) who combined models for COVID-19 growth rates. Granger and Newbold (1974) found combinations reduce error by 10-15% over single models in univariate series. Makridakis et al. (2018) highlight hybrids like neural networks and ARIMA outperforming individuals in M competitions.
Key Research Challenges
Weight Optimization
Determining optimal weights for combining forecasts remains computationally intensive. Khashei and Bijari (2010) show hybrid ARIMA-neural weights require cross-validation (786 citations). Performance degrades under non-stationarity.
Handling Uncertainty
Quantifying bias-variance trade-offs in ensembles under uncertainty is unresolved. Høyland and Wallace (2001) address scenario trees for multistage decisions but note approximation errors (676 citations). Morgan (2014) critiques expert elicitation biases in combinations.
Multi-Step Accuracy
Combining forecasts for multi-step horizons loses reliability. Ben Taieb et al. (2012) compare strategies in NN5 competition, finding recursive methods inferior (638 citations).
Essential Papers
Statistical and Machine Learning forecasting methods: Concerns and ways forward
Spyros Makridakis, Evangelos Spiliotis, Vassilios Assimakopoulos · 2018 · PLoS ONE · 1.3K citations
Machine Learning (ML) methods have been proposed in the academic literature as alternatives to statistical ones for time series forecasting. Yet, scant evidence is available about their relative pe...
Experience with Forecasting Univariate Time Series and the Combination of Forecasts
Paul Newbold, Clive W. J. Granger · 1974 · Journal of the Royal Statistical Society Series A (General) · 834 citations
A number of procedures for forecasting a time series from its own current and past values are surveyed. Forecasting performances of three methodsBox-Jenkins, Holt-Winters and stepwise autoregressio...
Autoregressive Distributed Lag (ARDL) cointegration technique: application and interpretation
Emeka Nkoro, Aham Kelvin Uko · 2016 · Journal of Statistical and Econometric Methods · 786 citations
Economic analysis suggests that there is a long run relationship between variables under consideration as stipulated by theory. This means that the long run relationship properties are intact. In o...
A novel hybridization of artificial neural networks and ARIMA models for time series forecasting
Mehdi Khashei, Mehdi Bijari · 2010 · Applied Soft Computing · 786 citations
Generating Scenario Trees for Multistage Decision Problems
Kjetil Høyland, Stein W. Wallace · 2001 · Management Science · 676 citations
In models of decision making under uncertainty we often are faced with the problem of representing the uncertainties in a form suitable for quantitative models. If the uncertainties are expressed i...
Use (and abuse) of expert elicitation in support of decision making for public policy
M. Granger Morgan · 2014 · Proceedings of the National Academy of Sciences · 663 citations
The elicitation of scientific and technical judgments from experts, in the form of subjective probability distributions, can be a valuable addition to other forms of evidence in support of public p...
A review and comparison of strategies for multi-step ahead time series forecasting based on the NN5 forecasting competition
Souhaib Ben Taieb, Gianluca Bontempi, Amir F. Atiya et al. · 2012 · Expert Systems with Applications · 638 citations
Reading Guide
Foundational Papers
Start with Newbold and Granger (1974) for empirical evidence on univariate combinations outperforming individuals; then Khashei and Bijari (2010) for ARIMA-neural hybrids.
Recent Advances
Makridakis et al. (2018) compares statistical-ML methods; Nikolopoulos et al. (2020) applies to COVID forecasting.
Core Methods
Simple/variance weighting (Granger 1974); stacking/hybrids (Khashei 2010); scenario trees (Høyland 2001).
How PapersFlow Helps You Research Forecast Combination Techniques
Discover & Search
Research Agent uses searchPapers('forecast combination techniques') to retrieve Granger and Newbold (1974), then citationGraph to map 834 citing works, and findSimilarPapers for weighting methods like Khashei and Bijari (2010). exaSearch uncovers hybrids in M4 competition data from Makridakis et al. (2018).
Analyze & Verify
Analysis Agent applies readPaperContent on Newbold and Granger (1974) to extract combination error metrics, verifyResponse with CoVe against M-competition benchmarks, and runPythonAnalysis to recompute weighted averages using NumPy/pandas on their datasets. GRADE scores evidence strength for bias reduction claims.
Synthesize & Write
Synthesis Agent detects gaps in multi-step combinations via Nikolopoulos et al. (2020), flags contradictions between ARIMA hybrids and ML in Makridakis et al. (2018); Writing Agent uses latexEditText for equations, latexSyncCitations for 10 papers, latexCompile for report, and exportMermaid for weighting flowchart diagrams.
Use Cases
"Reproduce Granger-Newbold forecast combination errors on economic series with Python"
Research Agent → searchPapers → Analysis Agent → runPythonAnalysis (NumPy/pandas on Box-Jenkins/Holt-Winters data) → matplotlib error plots and RMSE stats.
"Write LaTeX section comparing simple vs variance-weighted combinations citing 5 papers"
Synthesis Agent → gap detection → Writing Agent → latexEditText (weights equation) → latexSyncCitations (Granger 1974 et al.) → latexCompile → PDF with tables.
"Find GitHub repos implementing stacking forecast combinations from recent papers"
Research Agent → citationGraph (Makridakis 2018) → Code Discovery workflow (paperExtractUrls → paperFindGithubRepo → githubRepoInspect) → verified ARIMA-ML hybrid code.
Automated Workflows
Deep Research workflow scans 50+ papers via searchPapers on 'forecast combination', structures report with DeepScan's 7-step checkpoints verifying Granger (1974) results. Theorizer generates theory on optimal weighting from Makridakis et al. (2018) and Khashei (2010), using Chain-of-Verification for hallucination checks.
Frequently Asked Questions
What is forecast combination?
Forecast combination aggregates multiple model predictions, often via averaging or weighting, to reduce error.
What are main methods?
Simple averaging, variance-based weighting, and stacking; Granger and Newbold (1974) compared on univariate series.
What are key papers?
Granger and Newbold (1974, 834 citations) foundational; Makridakis et al. (2018, 1290 citations) on ML-statistical combos.
What are open problems?
Optimal dynamic weighting under uncertainty and multi-step degradation, per Ben Taieb et al. (2012).
Research Forecasting Techniques and Applications with AI
PapersFlow provides specialized AI tools for Decision Sciences researchers. Here are the most relevant for this topic:
Systematic Review
AI-powered evidence synthesis with documented search strategies
AI Literature Review
Automate paper discovery and synthesis across 474M+ papers
Deep Research Reports
Multi-source evidence synthesis with counter-evidence
See how researchers in Economics & Business use PapersFlow
Field-specific workflows, example queries, and use cases.
Start Researching Forecast Combination Techniques with AI
Search 474M+ papers, run AI-powered literature reviews, and write with integrated citations — all in one workspace.
See how PapersFlow works for Decision Sciences researchers