Transactions on Machine Learning and Data Mining (Print ISSN: 1865-6781) (ONLINE-ISSN: 2509-9337) (ISBN: 978-3-942952-73-6)
Volume 13 - Number 1 - July 2020 - Pages 3-25
Comparison of Multiple Linear Regression and Artificial Neural Network Models for the Prediction of Solid Waste Generation in Sri Lanka
C. L. Perera1 and M.G.N.A.S. Fernando2
1University of Colombo School of Computing, Colombo 07, Sri Lanka,
2University of Colombo School of Computing, Colombo 07, Sri Lanka
Abstract
In order to plan Solid Waste Management (SWM) in a sustainable way, reliable forecasting of generation of waste plays an important role. Predicting the quantity of generated waste is complex because it is affected by various influencing factors. In addition to population growth, economic development, household size, and employment changes would influence the solid waste generation interactively. The main objectives of this study are to identify significant factors influencing solid waste generation and to develop a model to predict solid waste generation in Sri Lanka. In this study, two predictive models, Multiple Linear Regression (MLR) Models and Artificial Neural Networks (ANN) were used. The MLR model which is a conventional method showed R2 values of 0.750, 0.544 and 0.769 for Biodegradable, Non-Biodegradable and Total waste, respectively. ANN model, a non-linear model showed R2 values of 0.846, 0.855 and 0.902 for Non-Biodegradable, Biodegradable and Total waste, respectively, which indicated higher predictive accuracy than MLR model. Therefore, in order to develop a prediction model with a higher predictive accuracy, ANN model is recommended.
Keywords:Solid Waste Management, Solid Waste Generation, Forecasting, Influencing Factors, Multiple Linear Regression, Artificial Neural Networks
Download Paper (379 KB)