Optimization of Machine Learning in Various Situations Using ICT-Based TVOC Sensors

Area of Science:

• Environmental Science
• Computer Science
• Chemical Engineering

Background:

• Total volatile organic compounds (TVOCs) pose significant risks to human health and ecosystems.
• Effective monitoring and reduction of TVOCs are crucial in industrial and laboratory settings.
• Existing artificial intelligence (AI) frameworks lack specificity regarding data characteristics and ground truth for diverse applications.

Purpose of the Study:

• To develop and evaluate an AI-driven computational framework for analyzing TVOC sensor data.
• To investigate the performance of various machine learning models in real-time TVOC detection.
• To enhance the accuracy and reliability of TVOC monitoring systems.

Main Methods:

• Utilized information and communications technology (ICT) semiconductor sensors for TVOC detection.
• Applied grounded data analysis and selected machine learning models, including Long Short-Term Memory (LSTM), Gated Recurrent Unit (GRU), and Recurrent Neural Network (RNN).
• Compared model performance against original sensory data, focusing on accuracy in diverse atmospheric conditions.

Main Results:

• Identified suitable analysis methods and machine learning models for TVOC sensing.
• Quantified the accuracy of LSTM, GRU, and RNN models in predicting TVOC concentrations.
• Demonstrated the potential for improved risk minimization in empirical applications.

Conclusions:

• The developed AI framework effectively utilizes ICT sensors and machine learning for accurate TVOC monitoring.
• The study provides a foundation for predicting abnormal situations and maintaining homeostasis in environments with TVOCs.
• This research is expected to significantly minimize risks in industrial and chemical applications through enhanced TVOC detection.