Combining data-intelligent algorithms for the assessment and predictive modeling of groundwater resources quality in parts of southeastern Nigeria.

Machine learning algorithms have proven useful in the estimation, classification, and prediction of water quality parameters. Similarly, indexical modeling has enhanced the evaluation and summarization of water quality. In Nigeria, works that have incorporated machine learning modeling in water quality analysis are scarce. Although studies across the globe have utilized overall index of pollution (OIP) and water quality index (WQI), works that have simulated and predicted them using machine learning algorithms seem to be scarce. Studies have not simulated nor predicted OIP. In this paper, several physicochemical parameters were analyzed and used for groundwater quality modeling in southeastern Nigeria based on integrated data-intelligent algorithms. Standard methods were followed in all the analysis and modeling performed in this work. OIP and WQI were computed, and their results revealed that 80% of the groundwater resources are suitable for drinking whereas 20% are highly polluted and unsuitable. Pearson's correlation analysis and R-mode hierarchical clustering revealed the possible sources of contamination. Meanwhile, agglomerative Q-mode hierarchical clustering and K-means (partitional) clustering were used to show the spatial demarcations of water quality in the area. Both clustering algorithms identified two main water quality classes-the suitable and unsuitable classes. Furthermore, multiple linear regression (MLR) model and multilayer perceptron neural networks (MLP-NN) were used for the estimation and prediction of the water quality indices. With low modeling errors, both MLR and MLP-NN showed very strong predictions, as their determination coefficient ranged between 0.999 and 1.000. However, MLR slightly outperformed the MLP-NN in the prediction of OIP. The findings of this paper would enhance sustainable water management in the study region and also contribute great insights to the national and global water quality prediction literatures.