Advancing waste-to-energy systems in institutional settings: Machine learning models for sustainable energy recovery...

Municipal solid waste (MSW) generation in university campuses across sub-Saharan Africa presents a critical management challenge and an underutilised energy recovery opportunity, compounded by structural information gaps including population undercounting and inadequate compositional data. This study applies an integrated machine learning (ML) framework to predict daily MSW generation, net calorific value (NCV), and electrical energy potential from primary empirical data collected at the University of Jos hostel complex, Nigeria. Four supervised regression algorithms - linear regression, random forests, gradient boosting, and a multi-layer perceptron neural network were trained and cross-validated on a 200-sample dataset constructed from 12-day empirical characterisation data (July 2021). Physical characterisation of 99.73 kg of hostel MSW identified food residue (35.50%) and polythene (32.44%) as the dominant fractions. The actual hostel population was field-estimated at 8,090 students 72.7% above official records yielding a per capita generation rate of 0.275 kg/cap/day and a total daily waste generation of 2.23 tons/day. Bomb calorimetry yielded a composite NCV of 15.23 MJ/kg, substantially exceeding the 7-8 MJ/kg viability threshold for thermal energy recovery. Linear regression achieved the highest MSW generation prediction accuracy (R² = 0.9361, RMSE = 52.95 kg/day, CV = 0.9224 ± 0.0083). A weighted ensemble model combining three algorithms achieved the best energy potential prediction (R² = 0.8286, RMSE = 133.80 kWh/day). The estimated monthly electrical potential of 2,776 kWh is sufficient to supply 2-4 individual campus facilities. Student population and food fraction were identified as the primary predictive features. The framework supports scenario-based planning under variable occupancy and enrollment conditions, and is directly transferable to other resource-constrained campuses across sub-Saharan Africa.