Classification and Identification of Contaminants in Recyclable Containers Based on a Recursive Feature...

Area of Science:

• Environmental Science
• Analytical Chemistry
• Computer Science

Background:

• Effective recycling mechanisms are crucial for environmental protection.
• Manual classification of recyclable containers is subjective and lacks accuracy.
• Automated methods are needed to improve the efficiency and reliability of container classification.

Purpose of the Study:

• To develop an automated identification method for recyclable containers using an electronic nose.
• To compare the effectiveness of Recursive Feature Elimination-Light Gradient Boosting Machine (RFE-LightGBM) with Principal Component Analysis (PCA) for feature dimensionality reduction.
• To enhance the classification accuracy of recyclable containers.

Main Methods:

• Odor features were extracted using an electronic nose.
• Feature datasets were constructed from gas response data.
• Dimensionality reduction was performed using both Principal Component Analysis (PCA) and Recursive Feature Elimination-Light Gradient Boosting Machine (RFE-LightGBM).
• Classification accuracy was evaluated on original, PCA-reduced, and RFE-LightGBM-reduced datasets.

Main Results:

• The RFE-LightGBM algorithm achieved the highest classification accuracy of 95%.
• The PCA dimensionality reduction dataset yielded a classification accuracy of 92.02%.
• The original feature dataset showed a classification accuracy of 88.38%.

Conclusions:

• The RFE-LightGBM algorithm offers a superior approach for automated classification of recyclable containers.
• Automated classification using electronic nose technology significantly enhances recycling efficiency.
• This method overcomes the limitations of manual classification subjectivity and improves accuracy.