Scholarlink

The basic modeling strategies of infectious diseases have become particularly important, especially during the times of the COVID-19 crisis. Although GNNs achieve remarkable accuracy in mimicking inter-regional interactions using spatio-temporal information, they infrequently capture the causal factors that govern the spread of epidemics. In an attempt to fill this research gap, this study puts forward the Multi-Relational Graph Attention Depth wise Separable Convolutional Neural Network with Satin Bowerbird Optimization Algorithm (MRGADSCNNet-SBOA). Both DSCNN and MRGAT components of the model are used to model temporal and spatial correlations of epidemic data. DSCNN can capture spatial interconnections well while adjusting node attributes through graph attention based on the relations that existed between neighboring areas even though it only has relatively few parameters for temporal relations. Here, the result reflects the model with the help of improved Satin Bowerbird Optimization Algorithm (SBOA) with the RMSE 1231, MAE 13823, MAPE 10.24%, PCC 96.42%, and CCC 98.91%. Due to its high reliability, the model offers a reasonable instrument for explaining epidemics in time and space.

错误：搜索内容不能为空，请输入英文关键词
错误：关键词超出字数限制，请精简
高级检索

Multi-relational Graph Attention-Based Depth Wise Separable Convolutional Neural Network for Spatio-Temporal Epidemic Forecasting

  • 会议论文

  • 发表日期：

    2026年

查看全文
  • Ravi Kumar Suggala,
  • B. Hema,
  • B. Naga Jahnavi,
  • Ch. Anannya Sai,
  • Ch. Saumya Prasanna,
  • J. Hari Keerthi

摘要

The basic modeling strategies of infectious diseases have become particularly important, especially during the times of the COVID-19 crisis. Although GNNs achieve remarkable accuracy in mimicking inter-regional interactions using spatio-temporal information, they infrequently capture the causal factors that govern the spread of epidemics. In an attempt to fill this research gap, this study puts forward the Multi-Relational Graph Attention Depth wise Separable Convolutional Neural Network with Satin Bowerbird Optimization Algorithm (MRGADSCNNet-SBOA). Both DSCNN and MRGAT components of the model are used to model temporal and spatial correlations of epidemic data. DSCNN can capture spatial interconnections well while adjusting node attributes through graph attention based on the relations that existed between neighboring areas even though it only has relatively few parameters for temporal relations. Here, the result reflects the model with the help of improved Satin Bowerbird Optimization Algorithm (SBOA) with the RMSE 1231, MAE 13823, MAPE 10.24%, PCC 96.42%, and CCC 98.91%. Due to its high reliability, the model offers a reasonable instrument for explaining epidemics in time and space.