Implementation of Data Mining to Support Waste Reduction Program in Special Region of Jakarta Using Time Series Algorithm and K-Means Clustering

Abstract

This study aims to analyze the trend of waste growth in Jakarta using the ARIMA method and to group areas based on waste volume using the K-Means Clustering algorithm. The waste accumulation problem at the Bantargebang TPST continues to worsen each year, with increasing volumes from various sub-districts. Data used in this study were obtained from the DKI Jakarta Environmental Agency, covering the period from January 2022 to April 2024, focusing on organic waste, plastic, and household hazardous waste (B3). The research applies the CRISP-DM methodology, consisting of business understanding, data understanding, data preparation, modeling, evaluation, and implementation. Data processing includes cleaning, normalization, and splitting into training and testing sets. The analysis results show that the ARIMA model achieves good forecasting accuracy, with MAPE, MAE, and RMSE values around 3652. The K-Means algorithm successfully classifies Jakarta areas into three main clusters dominated by organic, plastic, and mixed waste types. A web-based system was developed using Streamlit and MongoDB Atlas to facilitate data analysis and visualization for policymakers, especially the Environmental Agency. The study concludes that ARIMA is effective in forecasting waste growth, while K-Means supports more targeted waste management strategies. It is recommended to enhance the system by incorporating external variables such as policy changes and socio-economic factors, and to improve model accuracy using more advanced machine learning techniques. Additionally, the system should be continuously updated and expanded to support more optimal and sustainable waste management across Jakarta.