Scholarlink

<p>Urban waste-collection centres (WCCs) routinely overflow because maintenance routes are scheduled reactively rather than on data-driven forecasts. Overspill, odour, and leachate therefore threaten public health and sustainability targets in rapidly growing smart cities. We introduce ProWaste, an end-to-end Internet-of-Things and machine-learning platform that proactively prioritises WCC servicing. Fifteen automated and manual indicators, including population density, weather, maintenance history, and weekly waste build-up, are streamed from low-cost sensors, public APIs, and a mobile app to a cloud database. Twenty-five off-the-shelf classifiers were benchmarked under repeated stratified cross-validation; a Decision Tree Classifier offered the best balance of interpretability and near-top accuracy. Binary Particle Swarm Optimisation (BPSO) removed 80% of the inputs, revealing that three features alone predict criticality with&gt;99% accuracy on a hold-out test set. SHAP analysis confirms the interpretability of the three-feature model. The predicted class and confidence score are pushed to a Sustainable Smart Waste Management (SSWM) app that alerts field teams and dynamically reorders maintenance queues. Compared with current practice, ProWaste can eliminate missed pickups while reducing on-road inspections and data bandwidth. The proposed architecture is readily transferable to other cities and can be extended to recycling or composting streams.</p>

错误：搜索内容不能为空，请输入英文关键词
错误：关键词超出字数限制，请精简
高级检索

ProWaste for proactive urban waste management using IoT and machine learning

  • 期刊论文

  • 出版模式：

    开放获取

  • 发表日期：

    2025年7月30日

查看全文
  • Thompson Stephan,
  • S. M. Hari Krishna,
  • Chia-Chen Lin,
  • Ujjwal Sumesh,
  • Saurabh Agarwal,
  • Hyunsung Kim

摘要

Urban waste-collection centres (WCCs) routinely overflow because maintenance routes are scheduled reactively rather than on data-driven forecasts. Overspill, odour, and leachate therefore threaten public health and sustainability targets in rapidly growing smart cities. We introduce ProWaste, an end-to-end Internet-of-Things and machine-learning platform that proactively prioritises WCC servicing. Fifteen automated and manual indicators, including population density, weather, maintenance history, and weekly waste build-up, are streamed from low-cost sensors, public APIs, and a mobile app to a cloud database. Twenty-five off-the-shelf classifiers were benchmarked under repeated stratified cross-validation; a Decision Tree Classifier offered the best balance of interpretability and near-top accuracy. Binary Particle Swarm Optimisation (BPSO) removed 80% of the inputs, revealing that three features alone predict criticality with>99% accuracy on a hold-out test set. SHAP analysis confirms the interpretability of the three-feature model. The predicted class and confidence score are pushed to a Sustainable Smart Waste Management (SSWM) app that alerts field teams and dynamically reorders maintenance queues. Compared with current practice, ProWaste can eliminate missed pickups while reducing on-road inspections and data bandwidth. The proposed architecture is readily transferable to other cities and can be extended to recycling or composting streams.