Data-driven models applying in household hazardous waste: Amount prediction and classification in Shanghai

Area of Science:

• Environmental Science
• Data Science
• Waste Management

Background:

• Effective city management requires precise prediction and intelligent classification of household hazardous waste (HHW).
• Data-driven models offer potential but face challenges due to limited HHW data availability.
• Previous research has not fully explored advanced data-driven approaches for HHW forecasting and categorization.

Purpose of the Study:

• To develop and apply data-driven models for accurate HHW quantity prediction and classification in Shanghai.
• To utilize the prophet model for HHW forecasting and HVGGNet structures for HHW classification.
• To optimize the learning rate using cyclical learning rate method for efficient model training.

Main Methods:

• Prophet model integrated with Integration of Two Networks for HHW quantity forecasting.
• HVGGNet structures, based on VGG and transfer learning, for HHW classification.
• Cyclical learning rate method to optimize the global learning rate efficiently.

Main Results:

• The average HHW generation rate was 0.1 g/person/day.
• Key HHW categories include fluorescent lamps (30.6%), paint barrels (26.1%), medicine (26.2%), and batteries (15.8%).
• HVGGNet-11 achieved 90.5% precision for HHW sorting, and HHW in Shanghai is predicted to rise from 794.43 t in 2020 to 2049.67 t in 2025.

Conclusions:

• The study successfully demonstrated the efficacy of data-driven models for HHW management.
• HVGGNet-11 is highly suitable for intelligent HHW sorting.
• The predicted increase in HHW underscores the urgency for enhanced waste management infrastructure and policies.