Improving Gas Sensor Data Classification
Author Information
Author(s): Zeng Xianzhang, Shahzeb Muhammad, Cheng Xin, Shen Qiang, Xiao Hongyang, Xia Cao, Xia Yuanlin, Huang Yubo, Xu Jingfei, Wang Zhuqing, Nguyen Nam-Trung, Ngo Ha Duong
Primary Institution: Sichuan University
Hypothesis
Can machine learning techniques enhance the classification accuracy of gas sensor data with limited samples?
Conclusion
The study achieved a classification accuracy of 91.7% for gas types using a combination of PCA, SMOTE, and machine learning algorithms.
Supporting Evidence
- The study achieved 91.7% accuracy in differentiating gas types.
- PCA improved KNN and SVM classification, boosting AUC scores by 15.7% and 25.2%, respectively.
- SMOTE increased KNN’s accuracy by 2.1%, preserving data structure better than polynomial fitting.
- Combining PCA and SVM resulted in effective classification of gas samples into three types.
Takeaway
This study shows how to make gas sensors better at telling different gases apart, even when there's not much data to work with.
Methodology
The study used Principal Component Analysis (PCA) for dimensionality reduction, Synthetic Minority Over-sampling Technique (SMOTE) for data augmentation, and Support Vector Machine (SVM) and K-Nearest Neighbor (KNN) algorithms for classification.
Potential Biases
The classification models may favor majority categories due to the limited number of samples for minority categories.
Limitations
The original dataset had significant category imbalances, which could affect classification accuracy.
Digital Object Identifier (DOI)
Want to read the original?
Access the complete publication on the publisher's website