An Ensemble Learning Based Classification Approach for the Prediction of Household Solid Waste Generation

Area of Science:

• Environmental Science
• Computer Science
• Data Science

Background:

• Urbanization and smart city initiatives increase the complexity of waste management.
• Accurate solid waste generation prediction is crucial for efficient waste management planning.
• Existing machine learning models face challenges with limited datasets and features for waste prediction.

Purpose of the Study:

• To develop an ensemble learning technique for accurate weekly household waste generation prediction in urban areas.
• To address the challenges of limited datasets and features in waste generation forecasting.
• To improve upon existing state-of-the-art prediction models for solid waste management.

Main Methods:

• Developed an ensemble learning technique combining hyperparameter optimization (Optuna algorithm) and a meta-regressor model.
• Optimized individual machine learning models and used their outputs to train a meta linear regressor.
• Evaluated the ensemble model's performance against various single models and ensemble averages.

Main Results:

• The proposed ensemble method achieved an R2 score of 0.8 and a mean percentage error of 0.26.
• Outperformed state-of-the-art approaches including SARIMA, NARX, LightGBM, KNN, SVR, ETS, RF, XGBoosting, and ANN.
• Demonstrated superior performance compared to optimized single machine learning models and average ensemble results.

Conclusions:

• The ensemble learning technique significantly boosts prediction performance for household waste generation, even with feature-limited datasets.
• The findings offer practical implications for researchers and city authorities in optimizing waste management strategies.
• This approach provides a robust solution for accurate and efficient urban waste forecasting.