AI-Enabled Waste Classification as a Data-Driven Decision Support Tool for Circular Economy and Urban Sustainability

Efficient waste sorting is crucial for enabling circular-economy practices and resource recovery in smart cities. This paper evaluates both traditional machine-learning (Random Forest, SVM, AdaBoost) and deep-learning techniques including custom CNNs, VGG16, ResNet50, and three transfer-learning models (DenseNet121, EfficientNetB0, InceptionV3) for binary classification of 25 077 waste images (80/20 train/test split, augmented and resized to 150x150 px). The paper assesses the impact of Principal Component Analysis for dimensionality reduction on traditional models. DenseNet121 achieved the highest accuracy (91 %) and ROC-AUC (0.98), outperforming the best traditional classifier by 20 pp. Principal Component Analysis (PCA) showed negligible benefit for classical methods, whereas transfer learning substantially improved performance under limited-data conditions. Finally, we outline how these models integrate into a real-time Data-Driven Decision Support System for automated waste sorting, highlighting potential reductions in landfill use and lifecycle environmental impacts.)

💡 Research Summary

The paper addresses the growing challenge of urban solid‑waste management by developing an AI‑driven, data‑centric decision‑support system (DSS) for automated waste sorting, thereby supporting circular‑economy objectives. A publicly available dataset of 25,077 images, equally divided into organic and recyclable categories, was resized to 150 × 150 px and augmented through rotations, shifts, and flips. Traditional machine‑learning pipelines (Random Forest, Support Vector Machine, AdaBoost) were built on raw pixel vectors, with and without Principal Component Analysis (PCA) for dimensionality reduction. PCA yielded negligible performance gains, and the best traditional classifiers achieved only 70‑72 % accuracy.

In parallel, a suite of deep‑learning models was evaluated: a custom CNN, VGG16, ResNet‑50, and three transfer‑learning architectures—DenseNet‑121, EfficientNet‑B0, and Inception‑V3—initialized with ImageNet weights and fine‑tuned on the waste dataset. DenseNet‑121 emerged as the top performer, reaching 91 % accuracy and a ROC‑AUC of 0.98, a 20‑percentage‑point improvement over the best traditional method. Comprehensive metrics (precision, recall, F1‑score, ROC‑AUC) and confusion‑matrix analyses confirmed the superiority and robustness of the transfer‑learning approach, while early‑stopping prevented over‑fitting.

The authors propose a real‑time DSS architecture that streams images from smart bins or cameras, applies the selected model on edge devices, and feeds classification results back to automated sorting mechanisms. This pipeline promises reduced landfill dependency, higher recycling rates, and lower lifecycle carbon emissions.

Key contributions include (1) a systematic benchmark of classical versus deep models on a sizable binary waste dataset, (2) empirical evidence that PCA offers limited benefit for image‑based traditional classifiers, and (3) a practical integration blueprint for deploying the best model (DenseNet‑121) within smart‑city waste‑management infrastructure. Limitations are acknowledged: the binary classification scope, lack of multi‑class waste categories, absence of lightweight model evaluation for edge deployment, and no field‑level pilot validation. Future work should expand to multi‑class datasets, explore model compression/quantization, conduct real‑world pilot studies, and align the technology with policy and economic incentives to fully realize circular‑economy gains.