Jafarzade Naghmeh, Kisi Ozgur, Yousefi Mahmood, Baziar Mansour, Oskoei Vahide, Marufi Nilufar, Mohammadi Ali Akbar
Department of Natural Resources and Environment, Science and Research Branch, Islamic Azad University, Tehran, Iran.
Department of Civil Engineering, Technical University of Lübeck, 23562, Lübeck, Germany.
Heliyon. 2023 Jul 18;9(8):e18415. doi: 10.1016/j.heliyon.2023.e18415. eCollection 2023 Aug.
The Adaptive Neuro-Fuzzy Inference System (ANFIS) combines the strengths of both Artificial Neural Networks (ANNs) and Fuzzy Logic (FL) into a single framework. By doing so, it allows for quicker learning and adaptable interpretation capabilities, which are useful for modeling complex patterns and identifying nonlinear relationships. One significant challenge in assessing water quality is the difficulty and time-consuming nature of determining the various factors that impact it. Given this situation, predicting heavy metal levels in groundwater resources, both urban and rural, is essential. This paper investigates two methods, ANFIS-FCM and ANFIS-SUB, to determine their effectiveness in modeling Cadmium (Cd) in groundwater resources. The parameters to be considered are: dissolved solids (TDS), electroconductivity (EC), turbidity (TU), and pH were assumed to be the independent variables. A total of 51 sampling location were used with in the groundwater resource were used to develop the fuzzy models. For evaluating the performance of ANFIS-FCM and ANFIS-SUB models, three different performance criteria including the correlation coefficient, root mean square error, and sum square error were used for comparing the model outputs with actual outputs. Based on the obtained results from scatter plots of actual and predicted value by ANFIS-SUB and ANFIS- FCM models, the determination coefficient (R) value for total data, test and train sets is equal to 0.978, 0.982, 0.993 and to 0.983, 0.999 and 0.998 respectively. This result proved the Cd predictions of the implemented ANFIS-FCM model was significantly close to the measured all experimental data with R of 0.983. The performance of the implemented ANFIS-FCM model was compared with the ANFIS-SUB model and it is found that the ANFIS-FCM provided slightly higher accuracy than the ANFIS-SUB model. Also, the results obtained from the comparison between the predicted and the actual data indicated that the ANFIS-FCM and ANFIS-SUB have a strong potential in estimating the heavy metals in the groundwater with a high degree of accuracy.
自适应神经模糊推理系统（ANFIS）将人工神经网络（ANN）和模糊逻辑（FL）的优势结合在一个单一框架中。通过这样做，它实现了更快的学习速度和可适应的解释能力，这对于对复杂模式进行建模和识别非线性关系非常有用。评估水质的一个重大挑战是确定影响水质的各种因素既困难又耗时。鉴于这种情况，预测城乡地下水资源中的重金属含量至关重要。本文研究了两种方法，即ANFIS-FCM和ANFIS-SUB，以确定它们在对地下水资源中的镉（Cd）进行建模方面的有效性。所考虑的参数为：溶解固体（TDS）、电导率（EC）、浊度（TU）和pH值，假定它们为自变量。总共使用了地下水资源中的51个采样点来开发模糊模型。为了评估ANFIS-FCM和ANFIS-SUB模型的性能，使用了三个不同的性能标准，包括相关系数、均方根误差和平方和误差，用于将模型输出与实际输出进行比较。根据ANFIS-SUB和ANFIS-FCM模型的实际值与预测值的散点图所获得的结果，总数据、测试集和训练集的决定系数（R）值分别等于0.978、0.982、0.993以及0.983、0.999和0.998。这一结果证明，所实施的ANFIS-FCM模型对镉的预测与所有实验测量数据非常接近，R值为0.983。将所实施的ANFIS-FCM模型的性能与ANFIS-SUB模型进行了比较，发现ANFIS-FCM的准确性略高于ANFIS-SUB模型。此外，预测数据与实际数据比较所获得的结果表明，ANFIS-FCM和ANFIS-SUB在高精度估算地下水中的重金属方面具有很强潜力。
