Sequential Artificial Intelligence Models to Forecast Urban Solid Waste in the City of Sousse, Tunisia

Abstract

Urban solid waste (USW) is a major environmental problem for all countries. The rapid acceleration of economic growth, industrialization, urbanization, and population growth is correlated with the production of solid waste. Solid waste is usually disposed of in solid waste collection bins, so estimating the potential number of solid waste collection bins is necessary to plan an efficient USW management system. In this article, the purpose is to forecast the solid waste generation based on the monthly recorded amount of waste to determine the appropriate number of waste bins. Several artificial intelligence regression and classification approaches have been evaluated for their efficiency to estimate the number of bins. We highlight the effectiveness of sequential models, namely, long short-term memory (LSTM) and bidirectional LSTM (BLSTM), as waste data commonly consist of real-valued time series. Our experiments demonstrate the performance of the LSTM and BLSTM models, in terms of the number of waste bins prediction, when compared with other methods.