Published 2026-04-23
Dr. Elprida Agustinaresmi defended her dissertation in ITB’s Doctoral Program in Environmental Engineering on 22 April 2026. The study combined machine learning, spatial analysis, and socioeconomic variables to develop household-level priorities for plastic waste interventions across Bali.
Titled “Pengembangan Variabel Model Material Flow Sampah Plastik di Sumber dengan Analisis Geospasial dan Sosioekonomi”, the research addressed a gap in evidence about where and how households manage plastic waste. The source describes it as the first application of this combined approach at provincial scale in Indonesia.
Combining Household and Spatial Evidence
The study’s population comprised 905,935 households across Bali. Waste generation and composition were sampled at 200 homes for eight consecutive days under SNI 19-3964-1994. A survey of 700 respondents examined plastic waste handling and household socioeconomic conditions. Spatial data included geographic coordinates, building area, and remote-sensing information extracted for sampled homes.
A Light Gradient Boosting Machine regression model projected plastic waste generation, while a Random Forest classification model predicted handling behaviour, including selling, burning, and disposal into rivers. On test data, generation predictions achieved a reported R² of 0.88 and mean absolute percentage error of 12.88%.
The study reported that 26.72% of the projected population actively reduced plastic waste by selling it to collectors, while 48% burned plastic waste and 28% disposed of it into rivers. These behaviours are not presented as mutually exclusive categories.
Rural areas showed a distinctive pattern: households with moderate-to-low waste generation sold waste for economic reasons while also burning it or disposing of it into rivers. Economic motives and harmful practices could therefore coexist.
Location-Based Interventions
The research produced six location-based intervention levels, visualised in ArcGIS Pro, to help identify priorities for education, waste reduction infrastructure, and regulatory enforcement.
Multiple Correspondence Analysis and K-means clustering identified three risk-profile groups: households actively burning waste without reducing it, households with varied risks, and households with low-to-moderate risks. These distinctions support more targeted interventions.
The public defence took place in CIBE 306, chaired by Prof. Indah Rachmatiah Siti Salami and attended by supervisors Prof. Emenda Sembiring and Dr. Anjar Dimara Sakti, together with the examining team. The study offers Bali and other provinces a reference for evidence-based and spatially equitable plastic waste policy.