Proactive Fault Prediction of Fog Devices Using LSTM-CRP Conceptual Framework for IoT Applications.

Area of Science:

• Computer Science
• Artificial Intelligence
• Network Engineering

Background:

• The proliferation of real-time applications like IoT and video surveillance necessitates robust fog computing infrastructure.
• Fog devices face reliability challenges due to insufficient resources and harsh environments, impacting IoT service continuity.
• Scalable, proactive fault prediction is crucial for maintaining the reliability of fog devices and edge computing environments.

Purpose of the Study:

• To develop a proactive fault prediction method for fog devices experiencing resource limitations.
• To enhance the reliability and performance of fog nodes supporting Internet of Things (IoT) applications.
• To identify the precise causes of failure in fog devices due to inadequate resources.

Main Methods:

• A Recurrent Neural Network (RNN)-based approach integrating Long Short-Term Memory (LSTM) with a novel Computation Memory and Power (CRP) rule-based network policy.
• The CRP network policy is built upon the LSTM architecture to pinpoint failures stemming from resource inadequacy.
• Implementation of fault detectors and fault monitors within the framework to prevent fog node outages.

Main Results:

• The proposed LSTM and CRP method achieved 95.16% accuracy on training data and 98.69% on testing data.
• Demonstrated a significantly lower normalized root mean square error (0.017) for proactive fault prediction.
• Outperformed existing methods like traditional LSTM, Support Vector Machines (SVM), and Logistic Regression in prediction accuracy, speed, and resource management.

Conclusions:

• The LSTM and CRP network policy offers a highly accurate and efficient solution for proactive fault prediction in fog computing environments.
• The proposed framework effectively mitigates fog node failures caused by insufficient resources, ensuring uninterrupted IoT services.
• This approach represents a significant advancement in ensuring the reliability and performance of edge and fog computing systems.