Domestic Solid Waste Prediction with an Enhanced LSTM with SigmoReLU and RAdam Optimizer

A novel approach is presented to address the prediction challenge in domestic solid waste generation through the application of machine learning techniques. To overcome the limitations inherent in capturing intricate temporal patterns faced by conventional Long Short-Term Memory (LSTM) models designed for time series forecasting, an enhanced variant, termed e-LSTM, is introduced. This model incorporates crucial enhancements to rectify standard LSTM shortcomings. Introducing a hybrid activation function, SigmoRelu, bolsters the model’s capacity to grasp complex time series patterns. Furthermore, the RAdam optimizer is employed to optimize the learning process and improve convergence. Dropout layers are seamlessly integrated within the LSTM architecture to counter overfitting, ensuring robust generalization to novel data. A series of comprehensive experiments is conducted to compare the performance of the e-LSTM model against standard LSTM and GRU models, showcasing its noteworthy advancements. Notably, the e-LSTM model demonstrates superior predictive accuracy in forecasting waste generation compared to standard LSTM and GRU models. In essence, the proposed e-LSTM model represents a significant stride in domestic solid waste prediction, effectively mitigating the limitations of traditional LSTM models. The synergistic integration of SigmoRelu activation, RAdam optimization, and dropout mechanisms results in a resilient and accurate predictive framework. Empirical results affirm the model’s superiority, establishing it as a valuable tool for waste management applications and decision-making processes.