Enhancing e-waste management: a novel light gradient AdaBoost support vector classification approach.

Area of Science:

• Environmental Science
• Computer Science
• Machine Learning

Background:

• Electronic waste (e-waste) poses significant global environmental and health risks.
• Current e-waste management methods struggle to cover the entire product lifecycle.
• Accurate classification is crucial for effective e-waste recycling and disposal.

Purpose of the Study:

• To develop a novel algorithm for enhanced e-waste classification and management.
• To improve the accuracy of e-waste identification for better waste collection planning.
• To address limitations in traditional e-waste lifecycle management.

Main Methods:

• A framework incorporating data collection from diverse e-waste datasets.
• Image preprocessing techniques including scaling, rotation, flipping, noise removal, and label encoding.
• Feature extraction using modified Principal Component Analysis, followed by classification using Light Gradient Boosting Machine, AdaBoost, Support Vector Machine, and linear regression, with parameter tuning via dendritic growth search.

Main Results:

• The proposed model achieved high performance metrics: 98.1% precision, 96.1% recall, 97.1% F1-score, 98.5% accuracy, and 97.3% specificity.
• Demonstrated superior performance compared to existing e-waste classification models.
• Validated the effectiveness of the integrated machine learning approach for e-waste feature processing.

Conclusions:

• The developed algorithm significantly enhances e-waste classification accuracy.
• The framework offers a promising solution for optimizing e-waste management and collection.
• This approach contributes to mitigating environmental harm from improper e-waste disposal.