Coastal solid waste prediction by applying machine learning approaches (Case study: Noor, Mazandaran Province, Iran)

Nowadays, intelligent systems are used as innovative tools in different environmental issues. However, the prediction of short-term waste, unlike the long-term scale, is less developed due to more uncertainties and the difficulty in determining measurable independent parameters. In this study, two types of artificial neural networks (MLP and RBF) and two decision tree algorithms (CHAID and CART) have been used as effective tools for short-term forecasting of total waste production in coastal areas of Noor in Mazandaran Province, Iran. So that, average temperature, daily rainfall, sunny hours, maximum relative humidity and maximum wind speed were determined as the most important independent parameters, while the amount of waste produced in the coastal areas of Noor was considered as the dependent variable. Wastes from the coastal areas were gathered and their weights were analysed during 12 months from July 2017 through June 2018. Samplings were carried out twice a week, three weeks of a month and 12 months of a year, overall 72 times a year. The required meteorological data was gathered from the meteorological station in Noor. Then the sensitivity analysis was performed to check the independency of the major independent parameters. Thereafter, the mentioned machine learning approaches were applied to predict the short-term total waste production in IBM SPSS Modeler version 18 environment. In the applied models, 60% of data were used in training the model and the other 40% were used for model evaluation.  The results indicated that the CHAID tree algorithm exhibits a better performance in predicting total solid waste production compared to CART, MLP and RBF models. The mean absolute error and the correlation coefficient (R) of CHAID algorithm was 0.067 and 0.828, respectively.