Waste Prediction in Gross Pollutant Trap Using Machine Learning Approach

• Main Authors: Elpina, Sari, Tri Basuki, Kurniawan
• Format: Article
• Language: English
• Published: INTI International University 2023
• Subjects: QA Mathematics; QA76 Computer software; T Technology (General); TC Hydraulic engineering. Ocean engineering
• Online Access: http://eprints.intimal.edu.my/1800/1/jods2023_10.pdf; http://eprints.intimal.edu.my/1800/; http://ipublishing.intimal.edu.my/jods.html

Urbanization is often associated with decreased rainwater quality due to many factors, such as uncontrolled pollution and waste disposal. Therefore, managing water quality impacts in urban areas must be addressed to protect our environment. One of the maintenance steps is installing gross pollutant traps (GPT). The main objective of GPT is to remove dirty pollutants that are carried into the rainwater system before the rainwater enters the main river channel. At the same time, it is essential to understand that tropical climates are always associated with high rainfall intensity in a short period. It means that the amount of waste that GPT daily captures cannot be predicted well. It causes other problems, namely the emergence of difficulties in predicting the amount of waste that must be transported and moved from the GPT location to the final waste disposal site so that often, the rubbish that the GPT has caught will pile up at the GPT location without being able to be transported properly. Because the garbage vehicles that must transport the garbage are insufficient in number and capacity, it is necessary to have a model that can accurately predict the amount of waste that may be captured by each GPT based on past data on the amount of garbage that has been captured. This research compares 3 algorithms for predicting the amount of waste trapped by GPT: Simple Linear Regression, Multiple Linear Regression, and Polynomial Regression. The results show pretty good accuracy in our model, which is the RMSE is 1000. Next, a simple application was developed to lead the implementation of a load optimization scenario to show the importance of predicting the number of rubbish traps by each GPT by calculating how many trucks should be used to carry the garbage to the final waste disposal site.