An Enhanced Gas Sensor Data Classification Method Using Principal Component Analysis and Synthetic Minority...

Area of Science:

• Materials Science
• Sensor Technology
• Data Science

Background:

• Gas sensor data often suffers from high dimensionality and limited sample sizes, hindering accurate classification.
• Developing robust classification methods is crucial for reliable gas sensing applications.
• Existing methods may struggle with imbalanced datasets and complex feature spaces.

Purpose of the Study:

• To improve the classification accuracy of multi-dimensional and small datasets from gelatin-carbon black (CB-GE) composite film gas sensors.
• To evaluate the effectiveness of Principal Component Analysis (PCA) and Synthetic Minority Over-sampling Technique (SMOTE) in enhancing gas sensor data classification.
• To compare the performance of Support Vector Machine (SVM) and K-Nearest Neighbor (KNN) algorithms with and without data preprocessing techniques.

Main Methods:

• Utilized a gelatin-carbon black (CB-GE) composite film sensor for gas detection.
• Applied Principal Component Analysis (PCA) for dimensionality reduction of sensor data.
• Employed Synthetic Minority Over-sampling Technique (SMOTE) for data augmentation to address class imbalance.
• Implemented Support Vector Machine (SVM) and K-Nearest Neighbor (KNN) algorithms for gas type classification (ethanol, acetone, air).

Main Results:

• Achieved an overall classification accuracy of 91.7% for differentiating between ethanol, acetone, and air.
• PCA significantly improved classification performance, increasing Area Under the Curve (AUC) scores by 15.7% (SVM) and 25.2% (KNN).
• SMOTE enhanced KNN accuracy by 2.1% while better preserving data structure compared to polynomial fitting.

Conclusions:

• The combination of PCA and SMOTE offers a scalable and effective strategy for improving gas sensor data classification accuracy, particularly under data constraints.
• This approach demonstrates potential for expanding the reliability and effectiveness of gas sensors in diverse applications.
• The findings highlight the importance of data preprocessing techniques in maximizing the performance of machine learning algorithms for sensor data analysis.