Machine Learning Techniques for Predicting the Increase in Municipal Solid Waste Generation in Lima, Perú

Garcia-Sandoval S.-E.; Rodriguez C.R. · 2025 IEEE 4th International Conference on AI in Cybersecurity, ICAIC 2025 (2025) · DOI: 10.1109/icaic63015.2025.10848981

Resumen

According to data from the Ministry of the Environment (MINAM), Peru produces a daily average of 22,505.08 tons of municipal solid waste SW generated by its more than 33 million inhabitants. This is equivalent to about 0.8 kilograms of waste per person per day. However, only 1% of this waste is valorized; 53.5% goes to sanitary landfills, while 45.5% is in inadequate sites such as streets, riverbanks, landfills near homes, and environmental conservation areas. This poor management contributes to environmental contamination and represents a serious risk to public health. This study aims to develop a machine learning (ML) model to predict the increase of district solid waste in the department of Lima, Peru. The methodology uses a dataset that was elaborated with information provided by governmental entities covering 20142021. For the dataset processing, several ML models were applied, including multiple linear regression (MLR), Support Vector Machine (SVM), Gaussian Processes (GB), and Random Forest (RF). The analysis was performed using Python and its ML libraries, obtaining evaluation metrics such as the coefficient of determination (R2), mean square error (MSE), and mean absolute error (MAE). After comparing several algorithms, the results show that the RF model obtained the best performance, reaching an R2 of 95.0%, significantly outperforming the other models evaluated. The potential of these tools allows governmental entities to make decisions and apply strategies to prevent SW increase.

Instituciones: Universidad Nacional Mayor de San Marcos