Scholarlink

<p>Millions of children globally suffer from asthma, a common chronic respiratory disease that makes accurate and timely predictions essential to reducing health risks. This research proposes a sophisticated Deep Learning (DL) framework, AsthmaNet, designed for predicting asthma in children aged 0–9 years. The methodology encompasses advanced preprocessing, feature extraction, and hybrid optimization techniques to enhance predictive accuracy. The methodology includes an Iterative Imputer and Min-Max scaling in the preprocessing stage to improve the data quality. Moreover, feature extraction is performed to extract statistical, temporal, and quantile measures. A hybrid improved statistical tests, such as Chi-square and ANOVA, are used for determining the significant features in the feature selection phase. AsthmaNet is developed with five major DL architectures: Convolutional Neural Network (CNN), Recurrent Neural Network (RNN), Attention Mechanism, Transformer, and Graph Neural Network (GNN), while incorporating advanced design components such as multi-scale convolutional layers, residual connections, attention gates, and hybrid dilated convolution. A novel recommendation scheme based on Deep Reinforcement Learning (DRL) is incorporated to recommend specific treatment plans. The proposed framework is intended to increase the predictive precision and support healthcare practitioners in decision-making to increase positive impacts on patients and improve the quality of life for children with possible asthma. The outcomes of the designed technique are validated with existing models in terms of some error metrics. The developed model gained an MSE of 0.000158, an MAE of 0.012502, an MSLE of 0.000127, and an RMSE of 0.012585.</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

Asthmanet: integrating deep learning and reinforcement learning for pediatric asthma prediction and personalized recommendations

  • 期刊论文

  • 发表日期：

    2026年4月17日

查看全文
  • Pushkal Kumar Shukla,
  • Sarika Jain,
  • Siddharth Kalra

摘要

Millions of children globally suffer from asthma, a common chronic respiratory disease that makes accurate and timely predictions essential to reducing health risks. This research proposes a sophisticated Deep Learning (DL) framework, AsthmaNet, designed for predicting asthma in children aged 0–9 years. The methodology encompasses advanced preprocessing, feature extraction, and hybrid optimization techniques to enhance predictive accuracy. The methodology includes an Iterative Imputer and Min-Max scaling in the preprocessing stage to improve the data quality. Moreover, feature extraction is performed to extract statistical, temporal, and quantile measures. A hybrid improved statistical tests, such as Chi-square and ANOVA, are used for determining the significant features in the feature selection phase. AsthmaNet is developed with five major DL architectures: Convolutional Neural Network (CNN), Recurrent Neural Network (RNN), Attention Mechanism, Transformer, and Graph Neural Network (GNN), while incorporating advanced design components such as multi-scale convolutional layers, residual connections, attention gates, and hybrid dilated convolution. A novel recommendation scheme based on Deep Reinforcement Learning (DRL) is incorporated to recommend specific treatment plans. The proposed framework is intended to increase the predictive precision and support healthcare practitioners in decision-making to increase positive impacts on patients and improve the quality of life for children with possible asthma. The outcomes of the designed technique are validated with existing models in terms of some error metrics. The developed model gained an MSE of 0.000158, an MAE of 0.012502, an MSLE of 0.000127, and an RMSE of 0.012585.