Application of artificial neural network and multi-linear regression techniques in groundwater quality and health risk assessment around Egbema, Southeastern Nigeria

This paper examined the efficiency of artificial neural network (ANN) and multivariate linear regression (MLR) models in the prediction of groundwater quality parameters such as ecological risk index (ERI), pollution load index (PLI), metal pollution index (MPI), Nemerow pollution index (NPI), and geoaccumulation index (Igeo). 40 groundwater samples were collected systematically and analyzed for mainly heavy metals. Results revealed that adopting measured parameters is effective in modeling the parameters with high level of accuracy. Contamination factor results reveal that Ni, Zn, Pb, Cd, and Cu have relatively low values < 1 within the region while the Iron values ranged from low contamination to very high contamination (> 6). PLI, MPI, and ERI results indicated low pollution. NPI results indicated that the majority of the samples were heavily polluted. Quantification of Contamination results revealed that most of the sample's quality was geogenically influenced. Igeo results revealed that most of the samples had extreme pollution. The health risk assessment results revealed that children are substantially prone to more health risk more than adults. The ANN and MLR models showed a high effective tendency in the prediction of ERI, PLI, MPI, NPI and Igeo. Principal component analysis results showed appreciable variable loadings while the correlation matrix results reveal that there exists weak and positive correlation amongst elements. Based on the outcome of this study, this research recommends the use of ANN and MLR models in the prediction of groundwater quality parameters as they yielded positive, reliable, acceptable, and appropriate accuracy performances.


Introduction
Groundwater is a significant asset extremely pivotal for the supportability of human existence and the climate. It is trusted that around 65-75% of the earth's assets are water, access to quality water for drinking, homegrown and modern objects is restricted particularly in non-industrial nations (Eyankware et al. 2022a;Akakuru et al. 2022). Groundwater has become a fundamental requirement of life as more than 1.5 billion people around the world rely on groundwater as their primary source of drinking water (He et al. 2019;Agidi et al. 2022). Unfortunately, groundwater resources have faced serious problems in recent years, especially in urban, industrial and commercial areas. Chemicals used in agriculture, such as pesticides, make the groundwater that comes into contact with them undrinkable Opara et al. 2021). The restricted accessibility of 1 3 77 Page 2 of 20 value water, both in rustic and metropolitan regions is, for the most part, brought about by anthropogenic exercises. The anthropogenic elements which are famous wellsprings of water pollution framework emerge from modern, farming, and homegrown exercises to unfortunate waste administration (Sakram and Adimalla 2018;Eyankware and Akakuru 2022). Then again, the nature of groundwater generally relies upon the substance creation that is available and the focus levels of these synthetic boundaries are significantly gotten from the anthropogenic and land exercises in a specific district or region (Akakuru and Akudinobi 2018;Yahaya et al. 2021). Many elements control the groundwater quality; they incorporate geology, precipitation, mineral structure and dissolvability, oxidization, ionic trade, poor sterile circumstances, poor and uncontrolled composts, and pesticide application with minimal comprehension of substance cosmetics soils (Akakuru et al. 2017;Sakram and Adimalla 2018). Regardless of the overflow of groundwater, it is as yet unusable when its quality is significantly debased by substance defilements with expanding human populace, industrialization, urbanization and the resulting increment for the interest of water for both homegrown and modern purposes, the orderly expansion in the ramifications of dirtied water on man and the climate have been severally contemplated (Egbueri 2019; Barzegar et al. 2019;Akakuru et al. 2021bAkakuru et al. , 2022Eyankware et al. 2021). It is important to note that the concentration of heavy metals in groundwater resources has greatly increased due to anthropogenic activities and this poses threat worldwide. In the past 5-6 decades, there has been a high increase in the rate at which humans are exposed to heavy metals and this has been attributed to an exponential increase in heavy metals released from exploration, exploitation, mining, industrial, agricultural processes, technological advancement and her increase in the human population . Yang et al. (2019) has reported that two incidences among the top ten (10) global environmental disasters are attributed to contamination by heavy metals. Consequently, the release of large volumes of untreated effluents which contain a high concentration of heavy metals by industries into soils, surface and groundwater resources has further worsened the problem (Tziritis et al. 2017;Akakuru et al. 2021b). The toxic impacts of heavy metals often result from the intake of heavy metals, contaminated food, and water absorption through dermal contact with humans and inhalation of contaminated air. Thus, it is necessary and imperative that the sources, concentration and spread patterns of the heavy metals in surface and groundwater mediums are known; as such will help in sustainable environmental management programmes globally (Zhu et al 2020). Groundwater contamination has the characteristic of being difficult to find and control (Eyankware et al. 2020(Eyankware et al. , 2021Akakuru et al. 2021a). In any case, as of late, with the quick improvement of the general public and the sped-up advancement of urbanization, industrialization, and rural modernization, the effect of human activities has made groundwater defilement a serious level (Sakram and Adimalla 2018;Barzegar et al. 2019;Yahaya et al. 2021).
Human health risk assessment has been utilized to decide whether openness to substance, at any portion, could cause an expansion in the frequency of unfavourable impacts on human wellbeing (Li et al. 2015;Akakuru et al. 2022). Subsequently, people are concerned more about the connection between human wellbeing risk and groundwater contamination than contamination itself, so wellbeing risk evaluation is vital Eyankware et al. 2022c). Over the most recent couple of years, artificial neural networks (ANNs) have been generally applied in the space of water quality displaying. They are viewed as a forecasting device and have been generally utilized in different fields. They have additionally been utilized in hydrogeology to decide spring boundaries, assess the subjective attributes of groundwater (Hasda et al. 2020;Zhu et al. 2020;Mattas et al. 2019), and foresee groundwater levels (Ghorbanin et al. 2017;Nathan, et al. 2017;Singha et al. 2021). Regression models are best for laying out a relationship among dependent and independent variables, and they are thought of as the least difficult and generally direct type of model. They depend on the technique for least squares and are normally thought of as the main phase of an examination of the relationship among factors (Nathan et al. 2017;Zhu et al. 2020;Hasda et al. 2020).
Locally, several works have been done within the Egbema area in ascertaining the pollution level of groundwater but none has been done to integrate artificial neural networks and multi-linear regression models in the health risk assessment and quantification of contaminants in water resources around Egbema, Nigeria, hence this study which focuses on: 1. Integrating artificial neural networks and multi-linear regression models in the prediction of water quality parameters; 2. Assessing the groundwater quality using multivariate statistics such as principal component analysis (PCA) and correlation matrix; 3. Determining contamination and pollution status of groundwater resources using different models such as contamination factor (CF), pollution load index (PLI), Nemerow Pollution Index (NPI), metal pollution index (MPI), quantification of contamination (QoC), geoaccumulation index (Igeo) and ecological risk index (ERI); 4. To assess the health risk index of groundwater as it affects adults and children.

The study area
The review area is in the southwestern part of Imo and features the usual boundaries of Owerri to the east, Oguta to the north and Ogba/Egbema/Ndoni to the southwest of Rivers. In the 2006 assessment, the rating area was rated at 182,500. Occupants, however, as of late because of industrialization and urbanization, the review region has seen an extraordinary flood of populace deluge. The review region exists in Latitudes 4.9412 N to 4.9412 N and Longitudes 66.5382 E to 6.564 E. It covers an area of around 895 km 2 . The review region is to a great extent depleted by Urashi, Oloshi and Nkesa waterways and other occasional streams like Ugbujiagu and Nwagbedokpe streams. There are places in the review area that have important physiographic areas. The undulating wetland plain has something to do with its terrain. The lower land region is to a great extent underlain by the more youthful and approximately combined Benin Formation (Ezeigbo 1987).

Geology
Egbema is located in the eastern part of the Niger Delta Basin and is described by the Niger Delta lithography unit. Units include the Akada, Agbada, and Benin layers of age reduction requirements . The total width of the Tertiary debris is about 10,000 m, and all drill holes in the test area have evolved into a young Benin layer. Miocene-late developments generally consist of medium to coarse-grained sand units, with rocks properly placed at adjacent focal points of poorly consolidated sand and mud. The collaboration of sand and Earth in space shows a multi-spring framework. These sources are separated by a ground bond and their thickness determines the thickness of the aquifer in a particular well. Heavy rains in this area give the spring ample new energy (Aniwetalu and Akakuru 2015). The Niger Delta Basin otherwise called the Niger Delta territory, is an augmentation at the right bowl situated in the Niger Delta and bay of guinea on the detached mainland area close to the western bank of Nigeria with the thought or demonstrated admittance to Cameroon, Equatorial Guiana and Sao Tome and Principe. The bowl is extremely perplexing, and it conveys high monetary worth as it contains an exceptionally useful oil framework. The Niger Delta is one of the biggest subareas with around 75,000 km 2 , all-out areas of 300,000 km 2 , and a silt fill of 500,000 km 3 (Urom et al. 2021). The Niger Delta Basin was formed by crushed joints that were shocked during the division of the South American Plate, and the African Plate of the South Atlantic began to open. This shell collapse began in the late Jurassic and ended in the middle Cretaceous. As breaking proceeded with a few issues framed large numbers of the push blames likewise of this time syn-fracture sands and afterwards shale were stored in the late cretaceous (Ezeigbo 1987;Aigbadon et al 2022a, b;Onyekuru et al 2021).

Water sample collection and analysis
After assortment, the examples were housed in polypropylene measuring utensils for assessment. The example measuring utensils/holders to be used were very much scrubbed and inundated in refined water fermented with 1.0 ml of HNO 3 for three (3) days before the field inspecting exercise. A sum of 40 water tests was accumulated efficiently from different water sources across the review region (see Fig. 1). To acquire a delegate test that truly addresses the water assets, borehole tests were gathered following 5-10 min of siphoning. After rinsing the container with aliquots, each sample was strained into a sample bottle using a 0.45 m gauge disposable channel to ensure that all floating contaminants were completely removed. Field samples were treated with 1.0 ml concentrate to limit heavy metal precipitation. HNO3; Added 3 drops of HNO 3 using a new needle to avoid accretion. Purchased samples were properly stored at a temperature of 4° C in a tightly packed ice-filled container. On the way to the logic lab, the temperature was kept constant to prevent heat dissipation (Singh et al. 2005;Sehgal et al. 2012). Examples were synthetically analyzed for heavy metals Ni, Fe, Cu, Zn, Cd, and Pb using a fast continuous (FS) atomic absorption spectrophotometer (Varian 240 AA). All tests were performed according to the APHA specification (APHA 2012).

Artificial neural networks (ANNs)
ANN is used as a traditional insight-enhancing strategy and, as an extreme goal, contributes to the refinement and capacity of exploratory information and the transformation into structures of value that clients can process (Zhu et al. 2020;Mattas et al. 2019). A typical ANN consists of artificial processing elements called neurons or nodes connected by synapses. Neurons are collected in layers and data encoding is done in the most common way of preparation and learning. This structure is a widely involved model in hydrogeological applications with the ability to recognize designs between boundaries. The most capable exchange capabilities are the sigmoid logistic and hyperbolic tangent functions implemented in most ANN models (Hasda et al. 2020). Guided preparation depends on an "external educator" who gives objective value to each stage of preparation. The model understands how to change the synaptic load by thinking about the goal. The goal is to limit the error by searching for the ideal load (Ghorbanin et al. 2017;Nathan, et al. 2017;Singha et al. 2021). The standard measurable criterion used to assess ANN performance is mean squared error (MSE). This combines the predicted and ideal outcomes, as well as the confidence level (R 2 ). The current review used a computational ANN model of feedforward, management, and back-generation learning to predict groundwater ERI. The ANN model consisted of a six-component information layer (Ni, Fe, Zn, Pb, Cd, and Cu), a three-hub secret layer, and a result layer for which an EC rating was determined. Of the complete examples, 70% were used for setup and 30% were used for testing. Certain fake brain network structures were chosen in the process of using "experimental" techniques by changing the boundaries of information (number of neurons pushed in, number of hubs, percentage of pretest sets, etc.). Because I got the best presentation at.

Multiple linear regression (MLR)
MLR is considered to be a very valuable and accurate tool for establishing connections between reliable variables and various free factors considered indicators (Hasda et al. 2020). Various developers have effectively used this strategy in hydrogeology and hydrology to predict water quality and create factual models (Ghorbanin et al. 2017;Nathan et al. 2017). In this paper, a condition for assuming electrical conductivity using MLR have been created. Indicators were selected during the implementation of the Pearson relational coefficients because the selection of appropriate indicator coefficients is important for further developing the prediction level and narrowing down the required datasets (Singha et al. 2021). Connection factor (Pearson) is a fact-based tool commonly used to quantify and model the interrelationship and robustness design between two elements. MLR condition is introduced in condition 1.

Model performance
The performance of the model and the predictive power of the model were evaluated using R 2 . This emphasizes the magnitude of the percentage of variation in the y-axis variable explained by the x-axis variable. The range is 0 to 1. Mean Absolute Percentage (MAPE, %) is a measure of the prediction accuracy of the prediction method as defined in Eq. (2): where A t is the actual value, F t is the predicted value, and n is the number of samples. The root mean square error (RMSE) is the root mean square of the total error (Eq. (3)): where O i is the observed value, S i is the predicted value of the variable, and n is the number of observed values. Therefore, RMSE is a good measure of accuracy, but not between variables, but only to compare the prediction errors of different models or model configurations of a particular variable.

Contamination factor (CF)
CF in this study was determined by utilizing the Hakanson (1980) recipe where C k is the metal concentration and B k is the background/target level (Eyankware and Akakuru 2022).

Nemerow pollution index (NPI)
NPI PI Nemerow assess the overall level of soil contamination and take into account the levels of all heavy metals tested (Gong et al. 2008). Calculated using the following formula:

Metal pollution index (MPI)
The integrated effects of individual heavy metals on water quality were determined by MPI (Horton 1965;Eyankware and Akakuru 2022;Caeiro et al. 2005). Equation 7:

Quantification of contamination (QoC)
QoC has since been adopted as a contamination appraisal procedure integrated to decide the wellspring of a toxin (geogenic or anthropogenic).QoC was determined by the use of using Eq. 8

Geoaccumulation index (Igeo)
This is the rating level that made it possible to determine toxic substances by the Geoaccumulation Index (Igeo). According to Eyankware and Akakuru (2022), it is calculated as follows.
C n represents the content of toxic metal n, B n represents the background of toxic metal n, with 1.5 being a possible factor.

Potential ecological risk index (ERI)
ERI was first proposed by Hakanson (1980). where E i r represents the ERI of metal i th ; T i r is ith metal response factor to toxic. T i r of Zn, Cr, Pb, Cu, Ni, and Cd are 1, 2, 5, 5, 5 and 30, respectively (Islam et al. 2015;Mgbenu and Egberi 2019). The C i f is the contamination factor.

Evaluation of exposure
An important route of exposure of dissected trace components (Ni, Fe, Zn, Pb, Cd, and Cu) is usually by ingestion (Mgbenu and Egbueri 2019). Thus, the openness of the human, two child, and adult populations to these factors was measured using the chronic daily intake (CDI) given in Eqs. (1)... (12) (USEPA 1989; Mgbenu and Egbueri 2019). Where CDI (mg/kg/day); Cw stands for trace element concentration in water (mg/l). IRW is the water intake rate (IRW = 2L for adults, 1L for children). EF is the frequency of exposure (EF is usually equal to 365 days). ED is the duration of exposure (adult ED = 70 years, child ED = 6 years). BW is weight (adult BW = 70 kg; child BW = 15 kg); AT is average exposure time (adult AT = 25,550 days, child AT = 2190 days) (USEPA 1989; Mgbenu and Egbueri 2019).
The element's CDI was evaluated in terms of the Health Hazard Index (HQ) and the Health Hazard Index (HI). These are given by Eqs. (1), (13) and (14), respectively.

Multivariate analysis
Pearson's correlation coefficient was calculated using the Social Science Statistics Package (SPSS) version 17.0 and the PCA. R-Studio version 1.1456 software was used to model the ANN and MLR.

Results and discussion
The results of CF, PLI, MPI, ERI, and NPI are presented in Table 1.

CF
The CF reflects the pollution caused by focusing on the area. It shows a history of unique impurities in certain metals in biological media. The degree of heavy metal fixation at the base convergence of heavy metals was evaluated as a pollutant factor (Ogundele et al. 2017). CF has been used in groundwater studies to obtain the focal ratio of heavy metals to foundation values. The included model is used to show the benefits of the fouling factor. CF < 1, less fouling. 1 ≤ CF ≥ 3, moderate contamination; 3 ≤ CF ≥ 6, severe contamination; CF > 6, very high contamination (Bhutian et al. 2017;Akakuru et al. 2021a, b). The CF shown in Table 1 of this review indicates that Ni, Zn, Pb, Cd, and Cu generally have a low fixation of less than 1 over the entire range. Fe concentrations range from low impurities to very high impurities (> 6). Iron in rustic groundwater supplies is a typical issue: its focus level reaches from 0 to 50 mg/l, while the WHO suggested level is < 0.3 mg/l. The iron happens normally in the aquifer, however, levels in groundwater can be expanded by the disintegration of ferrous borehole and hand pump parts. Iron-bearing groundwater is much of the time perceptibly orange in variety, causing staining of clothing, and has an upsetting taste, which is obvious in drinking and food arrangement. Fe broke up in groundwater is in the decreased iron II structure. This structure is solvent and ordinarily creates no issues without help from anyone else. Iron II is oxidized to press III in contact with oxygen in the air or by the activity of iron-related microscopic organisms. Iron III structures insoluble hydroxides in water. These are corroded red and cause staining and blockage of screens, siphons, pipes, reticulation frameworks and so forth. If the iron hydroxide stores are created by iron microorganisms, they are likewise tacky and the issues of stain and blockage are ordinarily more awful. This finding is like that of Bhutian et al. (2017) in India and Nigeria (Yahaya et al. 2021).

PLI
PLI is an important tool for determining the level of heavy metal contamination in samples (Akakuru, et al. 2021a;Yang et al. 2019). PLI is usually categorized as uncontaminated (PLI < 1), moderately contaminated (PLI > 2), severely contaminated (2 < PLI > 3), or very severely contaminated (3 > PLI). According to the results (Table 1), the groundwater concentration value in the survey area was PLI < 1. Therefore, there is no pollution. This study contradicts a review attempted in India (Gopinath et al. 2019;Bhutian et al. 2017), which found out elevated PLI values in groundwater sources, but it is reliable to the work done in Nigeria by Yahaya et al. (2021). Figure 2 shows the spatial distribution of PLI within the region.

MPI
MPI is an important tool for mapping groundwater. Water with an MPI of 0.3 or less is said to be very pure, while  Figure 3 shows the distribution of MPI within the study area.

ERI
The ERI record was developed by Hakanson (1980), a Swedish researcher. It had been utilized to assess the antagonistic impacts of the impurities on the climate and humans and mirrors the harmfulness and natural awareness of the convergence of pollutants   Table 1 show that 82.5% of all samples are within the range of low ecological risk and 17.5% have moderate ecological risk. The results are consistent with the CF, PLI, and MPI performed in this study, indicating minimal exposure to heavy metals. Heavy metal pollution in groundwater addresses the threat to climate and food security due to the rapid development of industry and agriculture and the disruption of normal biological systems due to human pressures associated with population growth (Sarwar et al. 2017;Agidi et al. 2022;Akakuru et al. 2022). Ecological pollution and human openness to heavy metals stem from a variety of human practices, including mining, modern creation, and the use of metal-containing compounds in traditional and rural environments (Tchounwou et al. 2012). Figure 4 shows the spatial distribution of ERI in the region.

PNI
The Nemerow pollution index was used to understand the overall degree of pollution. This method provides a reasonable interpretation of the heavy metal load across each site, as different heavy metals can affect the site.  Table 1, it shows that 2.5% of the entire sample had a value less than 0.7, implying clean water, while 97.5% of the entire sample are > 3, implying that the samples are heavily polluted. Anthropogenic factors could be responsible for the high levels of pollution seen in this study. This finding contradicts a study conducted in Nigeria (Agidi 2022;Eyankware et al. 2022a, b, c). The spatial distribution of PNI within the region is presented in Fig. 5.  Table 2, whereas in Fe, 5% of the sample are positive indicating anthropogenic influence while 95% of the samples are negative implying geogenic influence. This means that the region's groundwater has been contaminated as a result of geogenic activity with very minor intercalation of anthropogenic sources. In any capacity, little is had significant awareness of the potential geogenic processes that might impact the movement of these metals in water resources. Heavy metal harmfulness has been a well established and developing ecological issue. A few heavy metals have been troubled by the dirt by geogenic sources like draining and enduring of rocks, as well as anthropogenic exercises like extreme metal mining, purifying, consuming petroleum derivatives, pesticide use, and sewage slime (Agidi et al. 2022). Whenever these metals enter the natural pecking order, in addition to the fact that an unfavourable effect on the plant has improved physiology, however, they additionally represent a constant and pestilence wellbeing chance to people. The discoveries of this study go against the discoveries of Egbueri (2019) in Ojoto, Nigeria.

Geoaccumulation Index (Igeo)
The geoaccumulation (Igeo) record was developed by to check the magnitude of pollution at significant concentrations of waste, water, debris, and soil, and is widely used to study global pollution status. Used (Haris et al. 2017). The properties of Igeo and their individual translations are Igeo ≤ 0 (essentially uncontaminated), 0 (uncontaminated), 1 (quite contaminated), 2 (moderately heavily contaminated), 3 (heavyly contaminated), 4 (obviously heavily contaminated) and Igeo ≥ 5 (clearly heavily contaminated) Very polluted), (Olujimi et al. 2014;Qing et al. 2015). From the aftereffects of the geoaccumulation file (Table 3), it revealed that 2.5% of the entire sample are unpolluted (< 1), 5% of the samples are within a moderate pollution area, 5% of the samples had strong pollution, 10% of the samples had moderate pollution, 75% of the samples had extreme pollution. This result agrees with the Nemerow pollution results but it is contrary to the result of the CF, PLI and MPI. Anthropogenic inputs could be responsible for the high levels of pollution seen in this study (Akakuru et al. 2021a, b;Agidi 2022;Eyankware et al. 2022a, b). The spatial distribution of Igeo within the region is presented in Fig. 6.

Health risk assessment
USEPA characterizes human wellbeing risk appraisal as the method involved with surveying unfriendly wellbeing and the probability of human openness to synthetic substances in a tainted climate. There are three potential openness courses that direct admission of drinking water, air inward breath, and skin retention. Since groundwater nitrogen doesn't volatilize, and skin retention is short of one-thousandth of drinking water admission, this article just considers openness through the drinking water consumption course. The measurement is communicated in the accompanying detailing. There are no growth-based temperature rules, but high water temperatures have been shown to be harmful to the human body and promote the development of microorganisms that can cause problems related to taste, aroma, skin protection and use (WHO 2017;Mgbenu and Egbueri 2019;Oli et al 2022). The results of the HI in Table 4 Table 4 Non-carcinogenic risk of heavy metals in terms of hazard quotient (HQ) and hazard index (HI) for children and adults  are substantially more presented with the constant exposure than the grown-ups Eyankware et al 2022c;Agidi et al. 2022).

Groundwater quality prediction by the ANN and MLR models
Having an accurate model that can predict the boundaries of interest is the key to saving the cost of verifying and assessing groundwater quality. This review demonstrates that both the MLR and ANN strategies are solid devices for calculating and expecting ERI, PLI, MPI, NPI, and Igeo. During the training and testing phase, 40 samples of chemical composition were used. Sensitivity analysis was performed on the ANN and MLR models to determine the relative weights of each input variable in making rational predictions. In the ANN foreseeing model, the determination of information factors is a critical stage, as it influences the model exhibition. In this unique situation, chose every one of the heavy metal boundaries was utilized as information factors for the forecast of ERI, PLI, MPI, NPI and Igeo. The predicted performance of the five ANN models was compared to the corresponding MLR models using R 2 , RMSE, and MAPE (quantitative indicators) (see Table 5). The results of this comparison are also shown in Table 5, and the optimized ANN model is shown in Fig. 7. Models with high R 2 values and significantly lower MAPE and RMSE values are considered optimal simulation models and are suitable for further analysis.    It is also important to point out that based on the R 2 values, the MLR appears to perform better as 60% of the values had an R2 value of 1.0 (ERI, NPI, and Igeo), implying that it is highly accurate in the prediction. The contribution level of the input parameters showing the predictor importance ( Fig. 8) are in this order: Fe > Pb > Cd > Ni > Cu > Zn (ERI), Pb > Cd > Fe > Cu > Zn > Ni (PLI and MPI), Fe > Pb > Cu > Ni > Zn > Cd (NPI and Igeo). This result indicates that the ANN and MLR models could be considered for further analysis since they showed a high effective tendency in the prediction of ERI, PLI, MPI, NPI and Igeo based on their respective performances.

PCA
PCA is a based arrangement technique that tries to make sense of the variety of countless interconnected factors (Eyankware, and Akakuru 2022;Akakuru et al. 2021a, b). It exhibits how factors are connected, which decreases the dataset's intricacy. PCA removes eigenvalues and eigenvectors from the first information's covariance framework. Head parts   Table 6). For PC3, there was no variable loading while PC4 has a loading of 16.7% among variables Zn (0.573). The aftereffect of this PCA affirms past outcomes and it further uncovers that the proceeded with anthropogenic exercises inside the territory has enormously impacted the dirt nearby groundwater tainting with heavy metals attracts a certified concern perspective on their unfortunate results for the living biota. The steady and non-biodegradable nature of profound metals works with their assortment in the environment. Provincial soils are getting a huge proportion of poisons from various sources. Past fundamental limits, heavy metals give a dangerous impact on human prosperity as they ruin the customary working of the living structures. The gigantic measure of waste made ought to be managed properly keeping into thought the normal assessments related to land treatment (Eyankware et al. 2021;Agidi et al. 2022;Akakuru et al. 2022).

Correlation Matrix
Connection grids are an important strategy used when evaluating connections between two elements. The relation coefficient is usually between -1 and + 1. Assuming the r-value is close to -1, the relationship is said to have a negative bias or vice versa. Whenever the value of r is close to + 1 the relationship is said to have a positive bias or is considered connected. If the value is zero, the centroid must be uncorrelated (Srivastava et al. 2014). The connection grid (Table 7) shows that there is a positive relationship between Pb and Fe (0.423). This result suggests a helpless and positive relationship between the components. This further means that salt water was not the main reactivated groundwater hotspot in the study area (Eyankware et al. 2021;Agidi et al. 2022;Akakuru et al. 2022). Lead and iron can enter drinking water when synthetic reactions occur in lead-containing plumbing materials. This is called consumption-melting or erosion of pipes and metal from pipes.

Conclusion
The artificial neural networks and multi-linear regression modelling of groundwater quality parameters and health risk assessment around Egbema, Southeastern Nigeria has been done. Contamination factor results reveal that Ni, Zn, Pb, Cd, and Cu have relatively low concentrations < 1 in the entire region while the Fe concentration range from low contamination to extremely high pollution (> 6). Pollution Load Index results showed no contamination exists. Metal Pollution Index results show that the groundwater is safe. Potential ecological risks assessment indicated the majority of the entire samples are within the low ecological risk area. Nemerow pollution results indicated that the majority of the samples were heavily polluted. Quantification of Contamination shows that majority of the sample's quality was geogenically influenced. Geoaccumulation index results showed that most of the samples had extreme pollution. The health risk assessment results revealed that children are substantially prone to health risks more than adults. The modelling results indicated that the ANN and MLR models could be considered for further analysis since they showed a highly effective and accurate tendency in the prediction of ERI, PLI, MPI, NPI and Igeo, based on their respective performances. Principal Component Analysis (PCA) results showed considerable variable loadings, while the correlation matrix results revealed that there exists weak and positive correlation amongst elements. This further implies that saltwater was not the major re-energised hotspot for the groundwater in the survey area. In light of the result of this study, it suggests the utilization of ANN and MLR modelling approaches in the prediction of groundwater quality parameters as they yielded positive and reliable performances.