A New Mixed-Gas-Detection Method Based on a Support Vector Machine Optimized by a Sparrow Search Algorithm.

Area of Science:

• Environmental Science
• Analytical Chemistry
• Machine Learning

Background:

• Traditional gas-concentration-prediction methods struggle with nonlinear data, leading to low accuracy.
• Recognizing mixed gases presents challenges due to overlapping signals and complex interactions.

Purpose of the Study:

• To develop an improved method for mixed-gas identification and concentration prediction.
• To address the limitations of existing methods in handling nonlinear data and low recognition rates.

Main Methods:

• Utilized Principal Component Analysis (PCA) for data dimensionality reduction.
• Employed a Sparrow Search Algorithm (SSA) to optimize Support Vector Machine (SVM) hyperparameters.
• Developed a hybrid SSA-SVM model for enhanced gas analysis.

Main Results:

• The SSA-SVM method demonstrated significantly improved classification accuracy for mixed-gas identification compared to Random Forest (RF), Extreme Learning Machine (ELM), and BP neural networks.
• Achieved a maximum fitting degree of 99.34% for single gas-concentration prediction and 97.55% for mixed-gas-concentration prediction.
• Experimental results confirmed the high recognition rate and concentration-prediction accuracy of the SSA-SVM method in gas-mixture detection.

Conclusions:

• The proposed SSA-SVM method offers a robust solution for accurate mixed-gas identification and concentration prediction.
• Optimizing SVM hyperparameters with SSA effectively enhances performance on nonlinear gas data.
• This approach shows promise for advanced gas-mixture detection systems.