Analysis of waste disposal and recycling processes in Türkiye with machine learning and artificial neural networks

Abstract

This study aims to analyze waste disposal and recovery data for 2022 and 2024 in Türkiye to reveal changes in national waste management and to make future predictions using multivariate machine learning and artificial neural network (ANN) models. Trends in disposal and recovery capacity were identified, and the main variables affecting waste processing volumes were modeled. Multiple linear regression and feedforward ANN models were created using the capacity and processing quantities obtained from Türkiye İstatistik Kurumu’s data. The ANN model was trained with three hidden layers and a 64-32-16 neuron structure, with a learning rate of 0.001. The dataset was divided into 80% training and 20% testing. The machine learning model’s test set accuracy coefficient was calculated as R² = 0.87, and the mean absolute error (MAE) was 1.96 million tons. The ANN model demonstrated even higher performance, reaching R² = 0.93 and an MAE of 1.41 million tons. ANNs accurately predicted total recycled waste to be 51.5 million tons in 2024, while the amount of landfilled waste was predicted with only a 3.2% margin of error. Model analyses revealed that increasing landfill capacity significantly increases disposal volume, while increasing the number of recycling facilities affects recovery volume through a strong, nonlinear relationship. The results show that ANN models can predict Türkiye’s waste management trends with high accuracy and prove that they provide a powerful decision support tool that can be used in future policy planning.