The amount of waste at landfills is increasing along with the growing population and human activities. Predicting the amount of waste has become one of the ways to address waste management issues. The quantity of upcoming waste can be determined through waste prediction, providing essential information for solving waste-related problems. This research involves modeling daily waste predictions using three machine learning algorithms: Linear Regression, Support Vector Regression, and Random Forest Regressor. The data used in this study is the waste data at Banyuurip landfill, Magelang City, covering the period from 2019 to 2022. In the data processing stage, attributes for data usage are selected, daily waste summation is performed, missing values are handled, and normalization is carried out using min-max normalization. The three machine learning algorithms are employed in the prediction modeling stage to obtain optimal parameters. The prediction model is evaluated by calculating the MSE. The results of the three waste prediction models using Linear Regression show a model with an MSE-train of 0.0086 and an MSE-test of 0.0083, while the RMSE-train is 0.0930 and the RMSE-test is 0.0915. The optimal SVR prediction model is obtained with hyperparameter combination C = 1, gamma = 1, and epsilon = 0.05, yielding MSE-train of 0.0030 and MSE-test of 0.0089, with RMSE-train at 0.0556 and RMSE-test at 0.0943. The Random Forest Regressor model results in a model with n_estimators of 500, random_state of 1, without the hyperparameter max_depth, and has MSE-train of 0.0012 and MSE-test of 0.0081, along with RMSE-train at 0.0353 and RMSE-test at 0.0901. Based on these three models, it is concluded that the best model is the Random Forest Regressor with the smallest MSE and RMSE values.
Copyrights © 2024