A machine learning approach for investigating the impact of seasonal variation on physical composition of municipal solid waste

The fault in the method of collection of waste-related data which has resulted in ill-defined and unreliable waste-data has necessitated the modeling approach. Despite the benefits of artificial-intelligence modeling in solving waste management problems in recent years, its applications in modeling the physical composition of waste, which is an important waste characteristic critical to sustainable waste management decisions are still lacking. In this study, an adaptive neuro-fuzzy inference system (ANFIS) model optimized with evolutionary algorithms, particle swarm optimization (PSO), and genetic algorithm (GA) was developed to investigate the effect of seasonal variation on the physical composition of solid waste using the city of Johannesburg as a case study. Three clustering techniques vis-à-vis grid partitioning (GP), subtractive clustering (SC), and fuzzy c-means (FCM) with varying combinations of their hyper-parameters were tested. The best models for each output are as follows: PSO-ANFIS-FCM with five clusters for organic waste (RMSE = 2.864), PSO-ANFIS-SC with cluster radius (CR) of 0.25 and squash factors (SF) of 1.3 for paper waste (RMSE = 2.543), GA-ANFIS-SC with CR of 0.35 and SF of 1.2 for plastic (RMSE = 4.329) and GA-ANFIS-GP with triangular membership function for textile (RMSE = 2.065). The result of this study revealed that both PSO-ANFIS and GA-ANFIS had a good performance in predicting the physical composition of waste stream with only a marginal variation.