Department of Statistics and Actuarial Science, School of Physical and Mathematical Sciences, University of Ghana, Legon, Accra, Ghana.
J Environ Public Health. 2021 Aug 14;2021:8622105. doi: 10.1155/2021/8622105. eCollection 2021.
Waste can be defined as solids or liquids unwanted by members of the society and meant to be disposed. In developing countries such as Ghana, the management of waste is the responsibility of the metropolitan authorities. These authorities do not seem to have effective management of the waste situation, and therefore, it is not unusual to see waste clog the drains and litter the streets of the capital city, Accra. The impact of waste on the environment, along with its associated health-related problems, cannot be overemphasized. The Joint Monitoring Programme report in 2015 ranked Ghana as the seventh dirtiest country in the world. The lack of effective waste management planning is evident in the large amount of waste dumped in open areas and gutters that remains uncollected. In planning for solid waste management, reliable data concerning waste generation, influencing factors on waste generation, and a reliable forecast of waste quantities are required. This study used two algorithms, namely, Levenberg-Marquardt and the Bayesian regularization, to estimate the parameters of an artificial neural network model fitted to predict the average monthly waste generated and critically assess the factors that influence solid waste generation in some selected districts of the Greater Accra region. The study found Bayesian regularization algorithm to be suitable with the minimum mean square error of 104.78559 on training data and 217.12465 on test data and higher correlation coefficients (0.99801 on training data, 0.99570 on test data, and 0.99767 on the overall data) between the target variables (average monthly waste generated) and the predicted outputs. House size, districts, employment category, dominant religion, and house type with respective importance of 0.56, 0.172, 0.061, 0.027, and 0.026 were found to be the top five important input variables required for forecasting household waste. It is recommended that efforts of the government and its stakeholders to reduce the amount of waste generated by households be directed at providing bins, increasing the frequency of waste collection (especially in highly populated areas), and managing the economic activities in the top five selected districts (Ledzekuku Krowor, Tema West, Asheidu Keteke, Ashaiman, and Ayawaso West), amongst others.
废物可以被定义为社会成员不需要并打算处理的固体或液体。在加纳等发展中国家，废物管理是大都市当局的责任。这些当局似乎对废物管理情况没有有效的管理，因此，看到废物堵塞排水道和首都阿克拉的街道上到处都是垃圾并不罕见。废物对环境的影响以及与其相关的健康问题怎么强调都不为过。2015 年联合监测计划报告将加纳列为世界上第七个最脏的国家。缺乏有效的废物管理规划体现在大量废物倾倒在未收集的露天区域和排水沟中。在固体废物管理规划中，需要有关废物产生、影响废物产生的因素以及废物数量可靠预测的可靠数据。本研究使用了两种算法，即列文伯格-马夸尔特算法和贝叶斯正则化算法，来估计人工神经网络模型的参数，该模型拟合来预测选定的阿克拉大都市区部分地区每月平均产生的废物量，并对影响固体废物产生的因素进行了严格评估。研究发现，贝叶斯正则化算法适合于训练数据的最小均方误差为 104.78559，测试数据的最小均方误差为 217.12465，以及较高的相关系数（训练数据为 0.99801，测试数据为 0.99570，总数据为 0.99767），在目标变量（每月平均产生的废物量）和预测输出之间。结果发现，房屋面积、地区、就业类别、主要宗教和房屋类型的重要性分别为 0.56、0.172、0.061、0.027 和 0.026，这是预测家庭废物所需的五个最重要的输入变量。建议政府及其利益相关者努力减少家庭产生的废物量，应集中精力提供垃圾桶、增加垃圾收集的频率（特别是在人口稠密的地区），并管理前五名选定地区（Ledzekuku Krowor、Tema West、Asheidu Keteke、Ashaiman 和 Ayawaso West）的经济活动等。
