Prediction of municipal solid waste generation using nonlinear autoregressive network

Area of Science:

• Environmental Science
• Data Science
• Waste Management

Background:

• Developing countries face significant challenges in solid waste management.
• Accurate prediction of waste generation is essential for effective strategic planning.
• Malaysia experiences a rapid increase in solid waste due to population growth and changing consumption patterns.

Purpose of the Study:

• To propose and evaluate an artificial neural network (ANN) model for predicting annual solid waste generation.
• To identify key demographic and economic variables influencing waste generation.
• To develop variable selection procedures for optimizing the predictive model.

Main Methods:

• Utilized a feedforward nonlinear autoregressive network with exogenous inputs (NARX) model.
• Incorporated demographic (population) and economic (GDP, employment, unemployment) variables as inputs.
• Employed variable selection techniques to identify significant predictors.
• Evaluated model performance using coefficient of determination (R(2)) and mean square error (MSE).

Main Results:

• The optimal ANN model achieved a high coefficient of determination (R(2)) of 0.97.
• The best model demonstrated a low testing mean square error (MSE) of 2.46.
• Key input variables identified were gross domestic product, population, and employment.

Conclusions:

• Artificial neural networks, specifically the NARX model, are effective for predicting solid waste generation in developing countries.
• Demographic and economic factors play a significant role in determining waste generation rates.
• The developed model provides a reliable tool for solid waste management planning.