Academic Journal
Προβολή σε EDS

Deep hybrid network with additive attention for accurate population forecasting.

Λεπτομέρειες βιβλιογραφικής εγγραφής
Τίτλος: Deep hybrid network with additive attention for accurate population forecasting.
Συγγραφείς: Yar, Hikmat1 (AUTHOR), Hussain, Adnan1 (AUTHOR), Khan, Zulfiqar Ahmad1,2 (AUTHOR), Kim, Min Je1 (AUTHOR), Baik, Sung Wook1 (AUTHOR) sbaik@sejong.ac.kr
Πηγή: Soft Computing - A Fusion of Foundations, Methodologies & Applications. Apr2026, Vol. 30 Issue 4, p2737-2754. 18p.
Θεματικοί όροι: *Population forecasting, *Machine learning, *Demography, *Artificial intelligence, *Convolutional neural networks, *Long short-term memory
Περίληψη: Population forecasting is crucial for informed decision-making and strategic planning in government and business, as it supports effective resource allocation, infrastructure development, and service provision to meet future demands. However, accurate population forecasting for small regions is challenging due to complex demographic processes, data sparsity, and the nonlinear nature of population growth. Despite advancements in time series forecasting methods, limited research has applied machine learning to small-area population forecasting, highlighting the need for models capable of capturing complex demographic trends to improve accuracy. To tackle these limitations, we propose a dual-stream architecture combining convolutional neural network (CNN) and bidirectional LSTM (BiLSTM) layers with additive attention mechanisms (Dual-CBA) for small-area population forecasting. The first stream utilizes CNN layers for spatial feature extraction, while the second stream employs a multilayered BiLSTM network to model temporal dependencies, capturing demographic trends and changes over time. The integration of additive attention allows the model to focus on relevant features in both spatial and temporal dimensions, enhancing its ability to learn complex demographic patterns and improve forecasting accuracy. The Dual-CBA is evaluated on five datasets namely Australia, New Zealand, Japan, United States, and South Korea. Results demonstrate that the proposed model achieves superior performance in terms of MAPE and MedAPE compared to the baselines. The model performance is also examined for 5- and 10-year population forecasts, showing optimal performance across these timeframes. Additionally, we also use the XAI method LIME to assess feature importance by region, providing insights into the key factors influencing population projections. [ABSTRACT FROM AUTHOR]
Βάση Δεδομένων: Academic Search Index
FullText Links:
  – Type: other
Text:
  Availability: 0
CustomLinks:
  – Url: https://dx.doi.org/doi:10.1007/s00500-025-10975-4
    Name: EDS - Springer Nature Journals (s7799221)
    Category: fullText
    Text: View record at Springer
Header DbId: asx
DbLabel: Academic Search Index
An: 192586155
RelevancyScore: 1431
AccessLevel: 6
PubType: Academic Journal
PubTypeId: academicJournal
PreciseRelevancyScore: 1430.76000976563
IllustrationInfo
Items – Name: Title
  Label: Title
  Group: Ti
  Data: Deep hybrid network with additive attention for accurate population forecasting.
– Name: Author
  Label: Authors
  Group: Au
  Data: <searchLink fieldCode="AR" term="%22Yar%2C+Hikmat%22">Yar, Hikmat</searchLink><relatesTo>1</relatesTo> (AUTHOR)<br /><searchLink fieldCode="AR" term="%22Hussain%2C+Adnan%22">Hussain, Adnan</searchLink><relatesTo>1</relatesTo> (AUTHOR)<br /><searchLink fieldCode="AR" term="%22Khan%2C+Zulfiqar+Ahmad%22">Khan, Zulfiqar Ahmad</searchLink><relatesTo>1,2</relatesTo> (AUTHOR)<br /><searchLink fieldCode="AR" term="%22Kim%2C+Min+Je%22">Kim, Min Je</searchLink><relatesTo>1</relatesTo> (AUTHOR)<br /><searchLink fieldCode="AR" term="%22Baik%2C+Sung+Wook%22">Baik, Sung Wook</searchLink><relatesTo>1</relatesTo> (AUTHOR)<i> sbaik@sejong.ac.kr</i>
– Name: TitleSource
  Label: Source
  Group: Src
  Data: <searchLink fieldCode="JN" term="%22Soft+Computing+-+A+Fusion+of+Foundations%2C+Methodologies+%26+Applications%22">Soft Computing - A Fusion of Foundations, Methodologies & Applications</searchLink>. Apr2026, Vol. 30 Issue 4, p2737-2754. 18p.
– Name: Subject
  Label: Subject Terms
  Group: Su
  Data: *<searchLink fieldCode="DE" term="%22Population+forecasting%22">Population forecasting</searchLink><br />*<searchLink fieldCode="DE" term="%22Machine+learning%22">Machine learning</searchLink><br />*<searchLink fieldCode="DE" term="%22Demography%22">Demography</searchLink><br />*<searchLink fieldCode="DE" term="%22Artificial+intelligence%22">Artificial intelligence</searchLink><br />*<searchLink fieldCode="DE" term="%22Convolutional+neural+networks%22">Convolutional neural networks</searchLink><br />*<searchLink fieldCode="DE" term="%22Long+short-term+memory%22">Long short-term memory</searchLink>
– Name: Abstract
  Label: Abstract
  Group: Ab
  Data: Population forecasting is crucial for informed decision-making and strategic planning in government and business, as it supports effective resource allocation, infrastructure development, and service provision to meet future demands. However, accurate population forecasting for small regions is challenging due to complex demographic processes, data sparsity, and the nonlinear nature of population growth. Despite advancements in time series forecasting methods, limited research has applied machine learning to small-area population forecasting, highlighting the need for models capable of capturing complex demographic trends to improve accuracy. To tackle these limitations, we propose a dual-stream architecture combining convolutional neural network (CNN) and bidirectional LSTM (BiLSTM) layers with additive attention mechanisms (Dual-CBA) for small-area population forecasting. The first stream utilizes CNN layers for spatial feature extraction, while the second stream employs a multilayered BiLSTM network to model temporal dependencies, capturing demographic trends and changes over time. The integration of additive attention allows the model to focus on relevant features in both spatial and temporal dimensions, enhancing its ability to learn complex demographic patterns and improve forecasting accuracy. The Dual-CBA is evaluated on five datasets namely Australia, New Zealand, Japan, United States, and South Korea. Results demonstrate that the proposed model achieves superior performance in terms of MAPE and MedAPE compared to the baselines. The model performance is also examined for 5- and 10-year population forecasts, showing optimal performance across these timeframes. Additionally, we also use the XAI method LIME to assess feature importance by region, providing insights into the key factors influencing population projections. [ABSTRACT FROM AUTHOR]
PLink https://search.ebscohost.com/login.aspx?direct=true&site=eds-live&db=asx&AN=192586155
RecordInfo BibRecord:
  BibEntity:
    Identifiers:
      – Type: doi
        Value: 10.1007/s00500-025-10975-4
    Languages:
      – Code: eng
        Text: English
    PhysicalDescription:
      Pagination:
        PageCount: 18
        StartPage: 2737
    Subjects:
      – SubjectFull: Population forecasting
        Type: general
      – SubjectFull: Machine learning
        Type: general
      – SubjectFull: Demography
        Type: general
      – SubjectFull: Artificial intelligence
        Type: general
      – SubjectFull: Convolutional neural networks
        Type: general
      – SubjectFull: Long short-term memory
        Type: general
    Titles:
      – TitleFull: Deep hybrid network with additive attention for accurate population forecasting.
        Type: main
  BibRelationships:
    HasContributorRelationships:
      – PersonEntity:
          Name:
            NameFull: Yar, Hikmat
      – PersonEntity:
          Name:
            NameFull: Hussain, Adnan
      – PersonEntity:
          Name:
            NameFull: Khan, Zulfiqar Ahmad
      – PersonEntity:
          Name:
            NameFull: Kim, Min Je
      – PersonEntity:
          Name:
            NameFull: Baik, Sung Wook
    IsPartOfRelationships:
      – BibEntity:
          Dates:
            – D: 15
              M: 04
              Text: Apr2026
              Type: published
              Y: 2026
          Identifiers:
            – Type: issn-print
              Value: 14327643
          Numbering:
            – Type: volume
              Value: 30
            – Type: issue
              Value: 4
          Titles:
            – TitleFull: Soft Computing - A Fusion of Foundations, Methodologies & Applications
              Type: main
ResultId 1