Statistical analysis of waste generation in Bukavu city using IoT sensor data

Abstract The city of Bukavu, in the Democratic Republic of Congo, faces a health and environmental crisis due to poor solid waste management, exacerbated by extreme population density (26,667 inhabitants per km²), rugged terrain (1,463-2,200 m), and inadequate infrastructure. This study proposes a statistical analysis of waste production using data from an intelligent Internet of Things (IoT) system. Stratified sampling and household monitoring sheets linked fill level measurements to household characteristics. After preprocessing, 47,013 measurements were analyzed. Using demographic weighting and an estimate of 220,255 households in 2026, the model based projection estimates a total daily production of 642 tonnes (234,320 tonnes per year), with an estimated weighted average of 2.92 kg per household per day, and per capita production projected to range from 0.26 to 0.58 kg per day depending on the municipality. ANOVA (p\textless{} 0.001) confirms significant differences: residential zones produce more (3.74 kg per household per day) than peri-urban zones (1.71 kg per household per day) and dense zones (2.36 kg per household per day). Waste density per kilometer of roadway shows critical disparities, notably in Kadutu (49.76 tonnes per km) and Ibanda (48.33 tonnes per km). A Random Forest model (R² = 0.257) identifies the three-day moving average as the most important predictor (57.3\%), highlighting autoregressive production patterns. K-means clustering revealed three user profiles: passive (47.4\%), standard (48.9\%), and active-critical (3.7\%). Monitoring detected 74 emptying events, with an average interval of 3 days. A projection to 2030, based on a 4.5\% annual growth rate, estimates 279,859 tonnes per year (more than 19.4\% compared to 2026). This study represents an in-depth statistical analysis of IoT data for waste management in a Sub-Saharan African city with such topographical and energy constraints. Results provide evidence based data to optimize collection routes and prioritize interventions.