Can machine learning improve small area population forecasts? A forecast combination approach

Grossman, Irina; Bandara, Kasun; Wilson, Tom; Kirley, Michael

doi:10.1016/j.compenvurbsys.2022.101806

Cited by 14 publications

(5 citation statements)

References 38 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…Overall, the ALL average model performed better than the STAT model, indicating that it was advantageous to include the GBMs in the ensemble. This is in line with previous work investigating combination methods for small area population point forecasts (Grossman et al, 2022).…”

Section: Discussionsupporting

confidence: 92%

“…The point forecast of the STAT model is created using the simple mean of the three statistical methods. Taking the mean of a set of forecasts has previously been shown to work well for small area population forecasts (Grossman et al, 2022). The methods to create the lower and upper forecasts are described below.…”

Section: Ensemble Modelsmentioning

confidence: 99%

“…We consider statistical and machine learning models, and combinations thereof, for Australian SA2 areas (Australian Bureau of Statistics, 2017a); these areas are the smallest units in the Australian Bureau of Statistics geographical hierarchy for which population statistics are regularly collected and published. This work will advance prior research by Grossman et al (2022) which investigated combination methods for Australian SA2s but focused exclusively on deterministic forecasts and included a slightly different set of individual models. Specifically, this paper has the following objectives:1.…”

Section: Introductionmentioning

confidence: 99%

See 2 more Smart Citations

Development and evaluation of probabilistic forecasting methods for small area populations

Grossman

Bandara

Wilson

et al. 2023

Environment and Planning B: Urban Analytics and City Science

Self Cite

View full text Add to dashboard Cite

Planning and development decisions in both the government and business sectors often require small area population forecasts. Unfortunately, current methods often produce forecasts that are inaccurate, particularly for remote areas and those with smaller populations. Such inaccuracy necessitates the development and evaluation of methods to forecast and communicate forecast uncertainty, however, little research has been conducted in this domain for small area populations. In this paper, we evaluate a set of probabilistic forecasting methods which include Autoregressive integrated moving average, Exponential Smoothing, THETA, LightGBM and XGBOOST, to produce point forecasts and 80% prediction intervals for Australian SA2 small area populations. We also investigate methods to combine the intervals to produce ensemble forecasts. Our results show that individual probabilistic methods generally produce prediction intervals which underestimate forecast uncertainty. Combining forecasts improves the overall accuracy of point forecasts and the coverage of their intervals, however, coverage still tends to be less than the expected 80% for all but the most conservative combination method.

show abstract

Section: Discussionsupporting

confidence: 92%

Section: Ensemble Modelsmentioning

confidence: 99%

Section: Introductionmentioning

confidence: 99%

See 1 more Smart Citation

Development and evaluation of probabilistic forecasting methods for small area populations

Grossman

Bandara

Wilson

et al. 2023

Environment and Planning B: Urban Analytics and City Science

Self Cite

View full text Add to dashboard Cite

show abstract

“…However, these are still in the development stage for even large area modelling, and there is some evidence for developing population estimates at the regional scale (Hu et al 2019). Extending and testing machine learning methods for small area population projections is an extremely recent development, and still in its early stages (Grossman et al 2022).…”

Section: Recent Developments In Small Area Population Projection In A...mentioning

confidence: 99%

Improving small area population projections

Baker,

Coffee,

Page

et al. 2024

AHURI Final Report

View full text Add to dashboard Cite

Related reports and documentsInquiry into projecting Australia's urban and regional futures: population dynamics, regional mobility and planning responses https://www.ahuri.edu.au/research-in-progress/Inquiry-intoprojecting-Australias-urban-and-regional-futures-populationdynamics-regional-mobility-and-planning-responses AHURI AHURI is a national independent research network with an expert not-for-profit research management company, AHURI Limited, at its centre. AHURI's mission is to deliver high quality research that influences policy development and practice change to improve the housing and urban environments of all Australians.Using high quality, independent evidence and through active, managed engagement, AHURI works to inform the policies and practices of governments and the housing and urban development industries, and stimulate debate in the broader Australian community. AHURI undertakes evidence-based policy development on a range of priority policy topics that are of interest to our audience groups, including housing and labour markets, urban growth and renewal, planning and infrastructure development, housing supply and affordability, homelessness, economic productivity, and social cohesion and wellbeing. AHURI Limited also gratefully acknowledges the contributions, both financial and in-kind, of its university research partners who have helped make the completion of this material possible.We also acknowledge the generous and insightful contributions of the industry and policy members of our Expert Panel. DisclaimerThe opinions in this report reflect the views of the authors and do not necessarily reflect those of AHURI Limited, its Board, its funding organisations or Inquiry Panel members. No responsibility is accepted by AHURI Limited, its Board or funders for the accuracy or omission of any statement, opinion, advice or information in this publication. AHURI journalAHURI Final Report journal series is a refereed series presenting the results of original research to a diverse readership of policy makers, researchers and practitioners. Peer review statementAn objective assessment of reports published in the AHURI journal series by carefully selected experts in the field ensures that material published is of the highest quality. The AHURI journal series employs a double-blind peer review of the full report, where anonymity is strictly observed between authors and referees.

show abstract

“…These algorithms are capable of utilizing vast amounts of data to construct models that can then make predictions based on past population data. Additionally, machine learning algorithms can develop time series models that enhance the accuracy of population predictions [5].…”

Section: Introductionmentioning

confidence: 99%

Forecasting The Population of China From 2020 To 2025 Based on Random Forest and Linear Regression

Wang

2024

HSET

View full text Add to dashboard Cite

The population dynamics of a country do play a vital role in its economic and political development. The COVID-19 epidemic has significantly affected the world's population. However, most people only care about the negative influence caused by COVID-19 but ignore the positive influence. This article uses two machine learning models, the random forest model, and the linear regression model, to predict the population change in China if there were no COVID-19 pandemic. With the predicted results, this article can compare the potential positive impact of the pandemic. This paper tries to fit two popular models, namely, Random Roest and Linear Regression to forecast the population of China from 2020 to 2025. The historical birth rate, death rate, and GDP growth rate of China are collected as features adding to the models to decline the error. For the Random Forest model, this paper set up an ensemble of decision trees to predict the future population of China. For the Linear Regression model, our features and population fit a linear connection. The findings of two models are compared in this article, and it is suggested that the random forest model is more suited for population forecasting. In addition, according to this study, the COVID-19 pandemic has some favorable effects on economic growth and birth rates. The article emphasizes the need to not only focus on the negative effects of the pandemic. Furthermore, the article points out that the linear regression model has poor fitting results for non-linear relationships. It suggests exploring more non-linear models for prediction and considering more influential parameters.

show abstract

Can machine learning improve small area population forecasts? A forecast combination approach

Cited by 14 publications

References 38 publications

Development and evaluation of probabilistic forecasting methods for small area populations

Development and evaluation of probabilistic forecasting methods for small area populations

Improving small area population projections

Forecasting The Population of China From 2020 To 2025 Based on Random Forest and Linear Regression

Contact Info

Product

Resources

About