Optimal Artificial Neural Network Type Selection Method for Usage in Smart House Systems

Area of Science:

• Artificial Intelligence
• Smart Home Technology
• Control Systems

Background:

• Smart home systems require processing complex, fuzzy input data from various sensors.
• Artificial neural networks (ANNs) are commonly used for this data processing.
• Different ANN architectures exhibit varying accuracies and capabilities for specific data types and control signal generation.

Purpose of the Study:

• To develop a method for selecting the optimal artificial neural network (ANN) type for smart home systems.
• To address the challenge of varying ANN performance with different data types and control signal generation.

Main Methods:

• Proposed an optimization problem to determine the best ANN type.
• Used the error of each ANN type in controlling a smart home subsystem as the optimization criterion.
• Trained different ANN types using identical historical input data.

Main Results:

• Identified dependencies between ANN types, network architecture (number of layers and neurons), and calculation errors.
• Demonstrated how network configuration impacts the accuracy of control signal generation for smart home subsystems.
• Quantified the error in parameter calculation relative to expected results for different ANNs.

Conclusions:

• The proposed optimization method effectively identifies the most suitable ANN for specific smart home applications.
• Understanding the relationship between ANN architecture and error is crucial for optimizing fuzzy data processing.
• This research provides a framework for enhancing the performance and reliability of smart home control systems.