Developing a Hybrid Regression-Metaheuristic Forecasting Model for University Solid Waste Generation

Abstract

The purpose of this study is to investigate and compare forecasting trends of university solid waste (USW) at a private university in Bangkok using a combination of statistical and metaheuristic algorithms. The university’s municipality solid waste data was prepared, collected, and converted so that it could be processed by a decision support system. Historical data is available for 16 years beginning with fiscal year 2005. Factors influencing USW quantities include the number of students, staffs, and others. USW is divided into two data sets: learning data sets and test data sets. The first will be examined using multiple regression and metaheuristics components. The learning datasets differ in data density because factor data is collected on an annual basis. As for the USW data, the data is recorded through the decision support system on a monthly basis. The second will be combined with the proposed method to determine the trend in substituting the prediction equation for USW management. The primary goal of this paper is to develop an effective USW forecasting model to deal with this problem by combining regression and metaheuristic techniques. Based on an empirical analysis of the indicators’ annual data and the USW’s monthly data from 2005 to 2020, we find that the hybrid method based on the linear equation performs well for all performance measures of the mean absolute percentage error (MAPE), mean absolute deviation (MAD), and mean square error (MSE). These findings may be useful in the preparation, configuration, and implementation of a waste management system at a university.