Deep convolutional neural network and IoT technology for healthcare

Area of Science:

• Artificial Intelligence
• Machine Learning
• Deep Learning
• Neural Networks
• Healthcare Informatics

Background:

• Deep Learning (DL) utilizes artificial neural networks (ANNs) to analyze complex data patterns, offering advantages over traditional machine learning algorithms.
• DL models consist of input, hidden, and output layers, enabling sophisticated data processing for accurate predictions.
• The research addresses the need for optimized DL architectures in healthcare applications.

Purpose of the Study:

• To determine the optimal number of hidden layers and activation function variations for neural networks in DL models.
• To analyze the effectiveness of different frameworks for building and comparing neural networks.
• To investigate techniques for accelerating neural network training without compromising accuracy, particularly for healthcare applications.

Main Methods:

• A dataset from Kaggle.com was utilized, focusing on reducing model layers for efficiency.
• The Rectified Linear Unit (ReLU) activation function was implemented with two fully connected layers.
• Model performance was evaluated using metrics such as R-squared (R2), Mean Squared Error (MSE), and Mean Absolute Error (MAE).

Main Results:

• A deep learning model trained with 19 features achieved an R2 of 0.89503 on the training set and 0.90707 on the test set.
• The optimized model demonstrated superior performance compared to a scikit-learn model, with specific metrics including MSE, RMSE, and MAE reported for both training and testing phases.
• The study confirmed that a dual hidden layer feed-forward neural network with the tanh activation function is effective for healthcare diagnostics.

Conclusions:

• Deep learning algorithms can enhance patient monitoring systems by providing timely health status updates and alerts.
• Integrating DL with the Internet of Things (IoT) facilitates automatic diagnosis and efficient data exchange in e-health systems.
• The research validates the selection of optimal neural network structures for dependable healthcare diagnostic systems.