Reliable prediction of municipal solid waste (MSW) generation rates is a significant element of planning and implementation of sustainable solid waste management strategies. In this study, the multi-layer perceptron artificial neural network (MLP-ANN) is applied to verify the prediction of annual generation rates of domestic, commercial and construction and demolition (C&D) wastes from the year 1997 to 2016 in Askar Landfill site in the Kingdom of Bahrain. The proposed robust predictive models incorporated selected explanatory variables to reflect the influence of social, demographical, economic, geographical and touristic factors upon waste generation rates (WGRs). The Mean Squared Error (MSE) and coefficient of determination (R2) are used as performance indicators to evaluate effectiveness of the developed models. MLP-ANN models exhibited strong accuracy in predictions with high R2 and low MSE values. The R2 values for domestic, commercial and C&D wastes are 0.95, 0.99 and 0.91, respectively. Our results show that the developed MLP-ANN models are effective for the prediction of WGRs from different sources and could be considered as a cost-effective approach for planning integrated MSW management systems.

中文翻译：

---

人工智能神经网络建模在生活、商业和建筑垃圾产生预测中的应用

对城市固体废物 (MSW) 生成率的可靠预测是规划和实施可持续固体废物管理战略的重要因素。在这项研究中，多层感知器人工神经网络 (MLP-ANN) 被用于验证对 Askar 垃圾填埋场 1997 年至 2016 年生活、商业和建筑和拆除 (C&D) 废物年生成率的预测。巴林王国。拟议的稳健预测模型结合了选定的解释变量，以反映社会、人口、经济、地理和旅游因素对废物产生率 (WGR) 的影响。均方误差 (MSE) 和决定系数 (R2) 用作性能指标来评估开发模型的有效性。MLP-ANN 模型在具有高 R2 和低 MSE 值的预测中表现出很强的准确性。生活垃圾、商业垃圾和 C&D 垃圾的 R2 值分别为 0.95、0.99 和 0.91。我们的结果表明，开发的 MLP-ANN 模型可有效预测来自不同来源的 WGR，并且可以被视为规划综合 MSW 管理系统的具有成本效益的方法。