Scholarlink

<p>The goal of this study is to construct a neuro computing-based intelligent framework with deep hidden layers for SIVR epidemic model which is divided into four different classes and employs the saturated incidence rate. The dataset of the SIVR epidemic model with saturate incidence rate model are generated using the reliable RK4 method. Stochastic optimization based on a dataset is designed to minimize mean square error by partitioning data 15% validation and testing while 70% training split. To show computational effectiveness of the model, five cases derived from the initial condition are used. Accuracy of the proposed scheme has been verified through outputs comparison. Best validation performance was achieved within the range <InlineEquation ID="IEq1"> <InlineMediaObject> <ImageObject Color="BlackWhite" FileRef="12190_2025_2639_Article_IEq1.gif" Format="GIF" Height="16" Rendition="HTML" Resolution="72" Type="Linedraw" Width="57" /> </InlineMediaObject> <EquationSource Format="TEX">\(\sim 10^{-06}\)</EquationSource> </InlineEquation> and 10<sup>−07</sup> providing an absolute error of 10<sup>−04</sup> to 10<sup>−06</sup> validating accuracy claim. Furthermore, some benchmark functions are simulated with the same DNN setup to show the performances of the models and the proposed methodology. This is the first time the SIVR model is approached with a deep neural network with tanh and ReLU in the hidden layers and Levenberg-Marquardt optimization-based deep surrogate model.</p>

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

Data-driven investigation of an epidemic model under saturated incidence rate using deep neural networks

  • 期刊论文

  • 发表日期：

    2025年9月3日

查看全文
  • Muhammad Farhan,
  • Waseem,
  • Adel Thaljaoui,
  • Mati ur Rahman

摘要

The goal of this study is to construct a neuro computing-based intelligent framework with deep hidden layers for SIVR epidemic model which is divided into four different classes and employs the saturated incidence rate. The dataset of the SIVR epidemic model with saturate incidence rate model are generated using the reliable RK4 method. Stochastic optimization based on a dataset is designed to minimize mean square error by partitioning data 15% validation and testing while 70% training split. To show computational effectiveness of the model, five cases derived from the initial condition are used. Accuracy of the proposed scheme has been verified through outputs comparison. Best validation performance was achieved within the range \(\sim 10^{-06}\) and 10−07 providing an absolute error of 10−04 to 10−06 validating accuracy claim. Furthermore, some benchmark functions are simulated with the same DNN setup to show the performances of the models and the proposed methodology. This is the first time the SIVR model is approached with a deep neural network with tanh and ReLU in the hidden layers and Levenberg-Marquardt optimization-based deep surrogate model.