Scholarlink

Diabetes is a long-term health issue, and early detection helps manage it better. This study proposes a deep learning model—Fully Connected Neural Network (FCNN)—for predicting diabetes using the Pima Indians Diabetes dataset. The FCNN uses Leaky ReLU, Adam optimizer, Batch Normalization, Dropout, and L2 weight decay to boost accuracy and prevent overfitting. Min-Max scaling was applied to normalize features, and a stratified train-test split ensured balanced data. Training used binary cross-entropy loss, a learning rate of 0.0005, and Early Stopping to avoid overfitting. The model reached 90% training accuracy and 78% testing accuracy, significantly improving over the original model’s 69%.

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

Deployment of Mobile Weather Monitoring Unit for Prediction and Forecasting by Integrating IoT and Machine Learning

  • 会议论文

  • 发表日期：

    2026年

查看全文
  • Amit Kumar Tiwari,
  • Humas Furquan,
  • Ankita Mishra,
  • Gyanshi Agrawal,
  • Jayesh Chauhan

摘要

Diabetes is a long-term health issue, and early detection helps manage it better. This study proposes a deep learning model—Fully Connected Neural Network (FCNN)—for predicting diabetes using the Pima Indians Diabetes dataset. The FCNN uses Leaky ReLU, Adam optimizer, Batch Normalization, Dropout, and L2 weight decay to boost accuracy and prevent overfitting. Min-Max scaling was applied to normalize features, and a stratified train-test split ensured balanced data. Training used binary cross-entropy loss, a learning rate of 0.0005, and Early Stopping to avoid overfitting. The model reached 90% training accuracy and 78% testing accuracy, significantly improving over the original model’s 69%.