<![CDATA[Inefficient waste collection routes result in significant operational costs and environmental impacts. Traditional static routes based on historical averages often deviate substantially from actual requirements. This study proposes an intelligent framework integrating Long Short-Term Memory (LSTM) networks for dynamic time-series forecasting with an Evolutionary Capacitated Vehicle Routing Problem (CVRP) optimizer. The LSTM model captures temporal waste generation patterns using a 7-day sliding window; these patterns are fed into a metaheuristic optimizer that minimizes travel distance to disposal sites while eliminating redundant trips. Experimental results demonstrate high prediction accuracy, with the Mean Squared Error (MSE) converging at 0.0001 during the validation phase. Furthermore, the optimization process achieved a 10.95% reduction (22.5 km/day) in average travel distance compared to the baseline model. In high-density scenarios, the framework improved route efficiency by up to 15.99%. The study concludes that combining deep learning memory capabilities with evolutionary optimization provides a reliable decision-support system for smart city waste management..]]>
Copyrights © 2026
