Feasibility of a Hybrid ANFIS-PSO Model to Predict Medical Waste: Case Study for Istanbul

Abstract

Accurate prediction of medical waste (MW) is critical for sustainable urban management. This study develops and validates a robust and reliable hybrid intelligent model for prediction of MW quantity. To reveal the effectiveness of the proposed model, a real case study focused on the city of Istanbul for MW is taken. First, a systematic variable selection process, incorporating Spearman Correlation and Variance Inflation Factor (VIF) analysis was employed to identify the most influential predictor variables. This process resulted in a final set of three key input variables including population density, literacy rate, and water consumption rate. Thus, to predict the MW amount in this study, a hybrid Adaptive Neuro-Fuzzy Inference System optimized by Particle Swarm Optimization (ANFIS-PSO) is proposed. The performance of this model was rigorously evaluated and benchmarked against four other machine learning (ML) models: a standard ANFIS, Support Vector Machine (SVM), Random Forest (RF), and an Artificial Neural Network (ANN). The results demonstrate that the proposed ANFIS-PSO model provides superior performance achieving the lowest error rates across all performance metrics. Accordingly, it yielded a Root Mean Square Error (RMSE) of 1837.75, a Mean Absolute Percentage Error (MAPE) of 5.60%, a Mean Absolute Error (MAE) of 1558.19 and Percent Bias (%PBIAS) of 2.04% on the test data. The findings confirm that the ANFIS-PSO hybrid model is a highly effective and useful tool for MW prediction offering a valuable resource for municipal authorities in sustainable waste management.