Robati Fateme Nazari, Akbarifard Hossein, Jalaee Seyyed Abdolmajid
Department of Economics, Faculty of Management and Economics, Shahid Bahonar University of Kerman, Kerman, Iran.
MethodsX. 2020 Oct 30;7:101120. doi: 10.1016/j.mex.2020.101120. eCollection 2020.
Poverty is a multifaceted phenomenon that its study and analysis from all dimensions requires accurate knowledge. In the past, poverty was measured only by the income approach. That is, only people's incomes were compared to the poverty line. But this approach does not identify other dimensions of poverty. Given the importance of discussing poverty in the economies of developing countries, this article examines and models poverty in the Islamic Republic of Iran. This article presents the internal and external dimensions of poverty in the period 1996-2017. In this paper, to model the poverty in Iran, the ANFIS method optimized with a differential evolution algorithm was used. In this method, a differential evolution algorithm was used to train the ANFIS system instead of the FIS system. To evaluate the strength of the model, mean squared error (MSE), root mean squared error (RMSE), Mean absolute error (MAE), STD_error, Mean_error criteria have been used. The data used in this paper are from the World Development Index (WDI) Database, the World Bank Good Governance Indices, the Heritage Foundation's Economic Freedom Indices, and the United Nations data. This information is related to Iran and in the period (1996-2017). The purpose of this paper is to train the ANFIS network with DE algorithm using time series data and to model the data related to the Iran Multidimensional Poverty Index using the trained network. Multidimensional Poverty Index is a very suitable index for monitoring data. Poverty is in society. With the help of this data, we can assess the trend of poverty and income distribution and welfare in this country. The results of this study showed that training the ANFIS system by differential evolution algorithm, can make a very good improvement in the modeling process and reduce error criteria and improve the accuracy of this method.•This article has been compiled with the aim of modeling poverty in the Islamic Republic of Iran.•The method used in this paper is ANFIS network training using the differential evolution algorithm•The use of evolutionary algorithms to train fuzzy systems and artificial neural networks leads to improved performance.
贫困是一个多方面的现象，从各个维度对其进行研究和分析需要准确的知识。过去，贫困仅通过收入方法来衡量。也就是说，只将人们的收入与贫困线进行比较。但这种方法并未识别出贫困的其他维度。鉴于在发展中国家经济中讨论贫困的重要性，本文对伊朗伊斯兰共和国的贫困情况进行了考察和建模。本文呈现了1996 - 2017年期间贫困的内部和外部维度。在本文中，为了对伊朗的贫困情况进行建模，使用了用差分进化算法优化的自适应神经模糊推理系统（ANFIS）方法。在这种方法中，使用差分进化算法来训练ANFIS系统而非模糊推理系统（FIS）。为了评估模型的强度，使用了均方误差（MSE）、均方根误差（RMSE）、平均绝对误差（MAE）、标准差误差（STD_error）、平均误差标准。本文使用的数据来自世界发展指标（WDI）数据库、世界银行善治指数、传统基金会的经济自由度指数以及联合国数据。这些信息与伊朗相关且处于1996 - 2017年期间。本文的目的是使用时间序列数据通过差分进化算法训练ANFIS网络，并使用训练好的网络对与伊朗多维贫困指数相关的数据进行建模。多维贫困指数是监测数据的一个非常合适的指数。贫困存在于社会中。借助这些数据，我们可以评估该国贫困、收入分配和福利的趋势。本研究结果表明，通过差分进化算法训练ANFIS系统，可以在建模过程中取得非常好的改进，降低误差标准并提高该方法的准确性。
•本文旨在对伊朗伊斯兰共和国的贫困情况进行建模。
•本文使用的方法是通过差分进化算法训练ANFIS网络
•使用进化算法训练模糊系统和人工神经网络可提高性能。
