1. Introduction

Rural environments are contaminated with diverse types of pollutants that affect water, soil, and air. Human activities, including urbanization and agricultural practices, contribute immensely to increasing pollutants in the environment, which might adversely affect ecosystems [1]. Major pollutants include emerging pollutants such as steroids, endocrine-disturbing compounds, pharmaceuticals, personal care products, artificial sweeteners, and surfactants. Mixtures of pollutants cause more adverse effects than individual compounds [2]. Industrialization, population growth, excessive use of pesticides and fertilizers, and leakage from water tanks are major sources of water pollution. These wastes have negative effects on human health [3, 4].

Additionally, stream water pollution caused by pesticides is a major environmental concern in farmed catchment areas. Many key factors contribute to this pollution, including weather, area topology, and crop management practices, all of which influence stream water quality [5, 6]. Thermal pollution damages water quality and ecosystem health [7]. It also causes the migration of creatures and can lead to the emergence of new species, disrupting reproduction and other biological processes [8, 9]. Furthermore, water pollution from plastics is significant, with about 280 million tons of plastic produced annually, much of which ends up in landfills and oceans [10]. Over 80% of sewage generated by human activities is discharged into rivers and oceans, resulting in environmental pollution and over fifty diseases [11]. Approximately 80% of diseases and 50% of child deaths worldwide are linked to poor water quality [12].

The health impacts of water pollution have been documented in the last several decades, including chronic poisoning, cancer related to microcystin, and health problems [13]. Water pollution has disastrous effects on infant mortality rates in peripheral and semi-peripheral countries [14, 15]. Long-term exposure to heavy metals may result in slowly progressing physical, muscular, and neurological degenerative processes that mimic Alzheimer's disease, Parkinson's disease, and muscular dystrophy [16]. With industrialization and urbanization, air pollution has become a life-threatening factor in many countries. Among air pollutants, the particulate matter with a diameter of less than 2.5 µm. It is a serious health problem because it might cause various respiratory and cardiovascular diseases [17, 18]. Worldwide, it was estimated that air pollution contributes to 800,000 premature deaths each year [19, 20]. The epidemiology and laboratory studies demonstrated that ambient air pollutants contributed to various respiratory problems, including bronchitis, emphysema, and asthma [3, 21, 22].

There was a notable spatial clustering effect on public health, with regional public health showing a convergence trend, and the adverse effects of air pollution’s negative externalities on public health were significant [17, 23]. Additionally, soil pollution affects soil, air, water quality, and people who live or work near polluted areas [24]. This type of pollution results from improper waste disposal, urbanization, agricultural chemicals, atmospheric decomposition, and soil erosion [25]. As urbanization increases, more people are exposed to environmental stressors, which can contribute to higher stress levels and compromised mental health [22, 26-29].

The associations between multiple environmental factors and self-assessed mental health for Chinese residents were significant [30]. Mental disorders have been associated with various aspects of anthropogenic changes to the environment, but the relative effects of different drivers were uncertain. Mental health disorders are generally linked to demographic and socioeconomic factors, and little is known about their interaction with the urban environment [30-32]. Contributions were made concerning both the perception of exposure to air pollution and the perception of the health effects associated with air pollution [33, 34]. Laboratory studies show that air pollutants can activate the neuroendocrine stress axis and modulate stress hormone levels, which could contribute to the development or exacerbation of psychological distress [31-37].

To predict and establish relationships between pollutants and mental health, artificial intelligence (AI) technology has become a necessity. It is widely used in variety of fields, including disease prediction, environmental monitoring, and pollutant prediction. There has also been an increase in the volume of research on air pollution using AI [38-43]. Perspectives of public understandings of global environmental risk have emphasized the interpretation and sense-making that takes place, modes of perception [44, 45], and the application of AI technologies and modules in studying the relationships between contaminants and mental health. For example, the water quality index was designed to describe various water quality variables using AI [46]. The impact of land use and management practices on stream water pollution was studied using modeling, simulation, and machine learning techniques to acquire knowledge about this complex domain. Results expressed the qualitative rules relating pollution factors to the temporal distribution of pesticide concentration [5].

Moreover, AI methods would enhance psychotherapy by providing therapists and patients with real or close-to-real-time recommendations [47, 48]. AI has been reported in various mental healthcare studies, where pertinent studies on the potential use of machine learning algorithms in assessing mental health [49]. Accordingly, the current study aimed to survey various types of pollutants in the rural environment, explore the perception of individuals on them, investigate the relationships between environmental pollutants and public health in these areas, and use artificial intelligence models in predicting the relationships among the presence of pollutants, mental health, and sustainability.

2.  Materials and methods

2.1. Study location

This study was conducted in Al-Kom Al-Akhdar Village, Hosh Eissa, El-Beheira Governorate, Egypt (Figure 1).

2.2. Experimental design and questionnaire

The experimental descriptive analytical model was considered for the current pilot study because of its importance and widespread use in scientific research. It is one of the scientific research methods that is capable of accurately analyzing the problem or phenomenon of scientific research and identifying the reasons for its occurrence. The current research project was characterized by broad comprehensive responses from urban participants through a pre-designed and validated survey. Data was collected using a face-to-face personal interview with respondents using a pre-designed questionnaire form that was prepared to achieve the objectives of the research. Questionnaire content validity was constructed following the face validity method. The questionnaire was pre-tested on three hundred people and necessary modifications were amended. Then it was pre-tested with the responses of thirty persons, with necessary modifications carried out on the used form  [50].

The questionnaire surveyed the demographic information: age, smoking, gender, marital status, possession of agriculture land, education level, occupation, volunteering with environmental protection agencies, suffering from pollution, culture, geographic openness, attendance of training courses related to environmental pollution, and sustainable development, sources of knowledge about environmental pollution and environmental sustainability, knowledge of environmental sustainability, environmental pollution and the relationship between pollution and individuals' mental health, global warming, and environmental pollutants, waste recycling is very important to reduce environmental pollution, public health and mental health and environmental pollutants, the most prevailing pollutants that affect the environment, the most common types of pollutants and psychological health, and how to reduce environmental pollution.

2.3. Data analytics using statistical analysis and machine learning classifiers

2.3.1. Simple statistics

The questionnaire results were statistically analyzed using descriptive methods such as arithmetic mean, standard deviation, percentages, and frequencies. Quantitative statistical methods such as Pearson's simple correlation coefficient, correlation analysis model, and ascending progressive multiple regression were employed to explain and interpret the results. The Statistical Package for Social Sciences (IBM SPSS), version 25, was used.

2.3.2. Machine learning classifiers

Machine learning classifiers are algorithms used in Artificial Intelligence -Supervised Learning to categorize or classify data into predefined categories or labels. These classifiers are designed to make predictions based on patterns and features extracted from input data [51]. Four classifiers were evaluated in the current study, including the support vector machine (SVM), which tries to find the hyperplane that best separates the two classes [46, 52]. SVM is a powerful algorithm that works well in high-dimensional spaces and is widely used in various applications, including image classification, bioinformatics, and natural language processing.  Logistic regression (LR) is a machine learning algorithm used for classification problems, where the goal is to predict binary class labels. It models the probability of the positive class as a function of the input variables using a logistic function, which maps any real-valued input to a probability between 0 and 1 [45]. Logistic regression uses maximum likelihood estimation (MLE) to learn the model parameters that maximize the likelihood of the observed data. Logistic regression is a simple yet effective algorithm that works well in many applications, including medical diagnosis, fraud detection, and credit scoring. Finally, the Random Forest algorithm (RF) is an ensemble learning algorithm that combines multiple decision trees to improve accuracy and reduce overfitting [41, 51, 53].  The algorithm involves two main steps: training and prediction. In the training step, a random subset of the training data was selected and for each subset, a set of features was randomly chosen. Then, a decision tree was built using the selected features and the subset of data. After that, the previous steps were repeated for a specified number of times (or until a stopping criterion was met). Finally, the decision trees were stored. In the prediction step, the new data points were run through each decision tree in the forest, and for each decision tree, the class of the data point was predicted based on the tree's decision rule. Then, the predictions from all trees were aggregated. After that, the majority vote for classification was done by taking the average of the predicted values for regression.

Evaluation matrices, also known as performance metrics, are used to assess the performance of machine learning models, algorithms, or systems. These metrics provide a quantitative measure of how well a model is performing in terms of accuracy (Equation 1). Precision: measures the proportion of true positives out of all positive predictions (Equation 2), recall: measures the proportion of true positives out of all actual positives (Equation 3), and F1 score, which is the harmonic mean of precision and recall and is a way to balance these two metrics (Equation 4). The choice of evaluation metrics depends on the specific problem and goals of the analysis. To evaluate the performance of the model that classifies current work, the following measures were used: Accuracy measures the percentage of correct predictions made by the model out of all predictions.

3. Results and discussions

Distinctive description of participants:  about 52.7% of participants were females, and 47.3% were males. The age of participants ranged from 18 to 70 years old, most of them (57.7%) were young (from 18 to 35 years) with an average of 38.8±15.8 years old (Table 1).

Table 1. The frequencies and percentages of gender, age, and spousal (social) status of participants in the current study (n=300).

Parameter	Frequency	Percentage
Gender
Male	142	47.3
Female	158	52.7
Total	300	100
Age
Young	173	57.7
Adult	94	31.3
Aged	33	11.0
Total	300	100
Marital Status
Married	204	68.0
Single	92	30.7
Widow	3	1.0
Divorced	1	0.3
Total	300	100

Source: Collected and calculated from the questionnaire conducted in the current study.

Table 2. Frequencies and percentages of occupation and possession of agricultural land of participants in the current research study.

Data in Table 3 showed the frequencies and percentages of smokers (nonsmokers, 63.0%), and smokers (37.0%), among participants. Also, 100% of participants suffer from one or more types of environmental pollutants.

Table 4. Frequencies and percentages of knowledge, cultural openness to others, and geographical openness to get information on pollution and psychological health (n=300).

Data in Table 4 showed that the categories of availability and use of knowledge sources were experienced by family and neighbors, television, radio, and agricultural extension engineers. The Ministry of Environment website, the agricultural extension magazine bulletins, the extension magazine, the extension meeting, agricultural researchers, university professors, scientific research articles, and social networking sites. The percentage of reading and seeking information sources was 1, 91, and 8% in the low, middle, and high categories, respectively. Also, data in Table 4 showed cultural openness, which is the extent of people’s culture, knowledge, and reading, following local and international news, using the internet for information and education, attending educational seminars at the university, learning about different cultures, and participating in exploratory trips for cultural development. These were classified into low, middle, and high, 1.3, 80.3, and 18.3%, respectively. Moreover, data on geographical openness showed that travel either outside the country, to the capital, to neighboring governorates' centers, and to the villages of the centers seeking knowledge on the environment and pollution. Data was categorized as low, middle, and high, with 6.3, 73.0, and 20.7%, respectively. Data in Table 5 collectively presented a description of participants' demographics, habits, education, and cultural interaction with information on pollutants and how they cope with them. Participants were mainly young (average age was 38±15 years), most of them were married, and worked in their own land. Educated persons were surveyed (12.63±4.78). Also, the table introduces the results of how they manifest pollution, cultural and geographical openness, learning through attending seminars, and training sessions that would impact their knowledge of environmental sustainability.

Table 5. Collective descriptive statistics of studied demographics, attitudes, and willingness to protect the environment from several types of pollutants.

The collected dataset preparation phase takes significant processing time; the preprocessing phase includes several steps, such as cleaning data, handling missing values, normalizing or scaling features, and encoding categorical variables. Each step helps improve the dataset so that the machine learning algorithms can interpret the data correctly and effectively.

Data in Figures 3, 4, and 5 after applying the machine learning models showed the pollutants that mostly affected the environment arranged in descending order from most influential to least influential as follows dumping of solid waste, agricultural and household waste and not to be recycled, spaying agriculture pesticide, exhausts from vehicles, trains, ships and aircraft, rabid urbanization, overfishing, deforestation and trees, plastic consumption, fossil fuel combustion and steady population growth. The results provided describe an experimental assessment of pollutants affecting the environment, with an arrangement in descending order from the most influential to the least influential.

The most common types of pollution found in the environment (Figure 6) were air pollution (12%), water pollution (12%), agricultural soil pollutants (14%), noise (12%), and visual pollution (12%). The provided results present the distribution of several types of pollution in the environment, with each type representing a percentage of the total pollution. The 12% allocation to water pollution suggests that a portion of the total pollution affects water bodies such as rivers, lakes, and oceans. Agricultural soil pollution accounts for 14% of the overall pollution. This type of pollution is associated with the introduction of contaminants into the soil through agricultural practices. Pesticides, fertilizers, and chemicals used in farming can contribute to soil pollution, affecting soil health, and potentially affecting crops and ecosystems. Noise pollution, being 12% of the total pollution, involves the presence of unwanted or harmful sounds in the environment. Visual pollution, accounting for 12% of the total, refers to the presence of unsightly or visually intrusive elements in the environment. This can include factors such as litter, poorly maintained structures, and other visual disturbances.

In Figure 7, the sources of air pollution include kiln smoke in 12%, chemical industries (foundries) and chemical fertilizers in 14%, and fires and smoke in 13%. The results presented in the figure outline the sources of air pollution and their respective contributions in terms of percentages. Kiln smoke represents 12% of the total air pollution. Kilns are furnaces or ovens used for various industrial processes such as the production of ceramics, cement, and lime. The combustion of fuels in kilns can release pollutants into the air, including particulate matter, sulfur dioxide, and other harmful substances. The 12% allocation suggests that this specific source significantly contributes to overall air pollution.

Chemical industries, including foundries, and the use of chemical fertilizers collectively contribute to 14% of air pollution. Foundries are industrial facilities involved in metal casting, and their operations can release pollutants into the air, such as metal dust and emissions from metal smelting. Chemical fertilizers, when applied to agricultural fields, can release ammonia and nitrogen oxides into the air. The combined 14% highlights the impact of industrial and agricultural activities on air quality. Fires and smoke contribute to 13% of air pollution.

Common water pollutants (Figure 8) were residues of factories and foundries that are thrown into water sources (26%), leakage of ammonia gas from human waste (21%), industrial detergents (26%), and dumping of dead animals and their remains in freshwater sources (27%). It's important to note that these results highlight anthropogenic sources of water pollution, underscoring the importance of proper waste management, industrial practices, and pollution prevention strategies to protect and preserve water quality. Collectively, Figure 9 shows the correlation heatmap of main factors affecting the relation between presence of pollutants and mental health of rural inhabitants.

Moreover, the SVM classified and predicted the mental health outcomes based on exposure to pollutants. Relevant features that might affect mental health outcomes include specific pollutants, demographic information, or other relevant variables. The model was evaluated using proper metrics (accuracy, precision, recall, and F1-score) to generalize for unseen data. It was clear from the data in Figure 10 A that the accuracy rate, precision, recall, and F1-score of the SVM were 70, 49, 70, and 58%, respectively. Also, decision tree classifier provided interpretability of the understanding of factors that influence mental health outcomes (Figure 10 B). Decision tree classifier proceeded with relevant features that might impact mental health outcomes, including specific pollutants and demographic information, using the decision tree model on the testing set, using appropriate metrics (accuracy, precision, recall, and F1-score with 62, 57, 61, and 58%, respectively.

Random Forest classifier studied the relationship between mental health and pollutants and offered robust insights. Figure 10 C showed that the accuracy rate of the random forest classifier was 65%, the precision was almost 54%, the recall was 65% and F1-score was almost 57%. Logistic regression tree classifier: Logistic regression tree (LRT) is in the context of a decision tree-based ensemble model, like a Gradient Boosting Machine (GBM), or an Adaptive Boosting (AdaBoost) classifier for studying mental health and pollutants (Figure 10 D). The accuracy rate of the logistic regression forest classifier was 70%, the precision was 49%, the recall was 70%, and F1-score was almost 57%. Collectively, the four classifiers were compared in Figure 10. Based on the earlier results, it could be concluded that the best classifiers in terms of accuracy were SVM and logistic regression. Logistic regression and SVM prevailed according to recall percentages. Finally, the decision tree classifier was the best choice in terms of F1-score.

Random forest classifier: using a Random Forest classifier to study the relationship between mental health and pollutants offered robust predictions and insights. Random forest model is an ensemble learning method that builds multiple decision trees and merges their predictions. Figure 15 showed that the accuracy rate of the random forest classifier was 65%, the precision was almost 54%, the recall was 65% and F1-score was almost 57%.

Logistic regression tree classifier: Logistic regression tree (LRT) is in the context of a decision tree-based ensemble model, like a Gradient Boosting Machine (GBM), or an Adaptive Boosting (AdaBoost) classifier for studying mental health and pollutants. The accuracy rate of the logistic regression forest classifier was 70%, the precision was 49%, the recall was 70%, and F1-score was almost 57% (Figure 16). Collectively, the four classifiers were compared in Figure 16. Based on the earlier results, it could be concluded that the best classifiers in terms of accuracy were SVM and logistic regression. Logistic regression and SVM prevailed according to recall percentages. Finally, the decision tree classifier was the best choice in terms of F1-score.

Table 17. Suggestions and solutions for preserving good mental health in relation to environmental pollutants.

Data in Table 17 showed the suggestions and solutions for preserving good mental health in relation to environmental pollutants from the point of view of the citizens surveyed. Some solutions at the individual level to reduce the presence of pollutants were: rationalizing the use of environmental resources and not wasting them, recycling household waste to create useful by-products, and not throwing dead animals into waterways.

In Table 18, some of the suggested solutions are provided by participants to reach sustainability and environmental protection. Research results showed suggested views arranged from the most popular to the least as the following: increase the environmental awareness among children from a young age in schools (100%), cash in return of waste recycling (97%), reduction of greenhouse gas emissions (96%), use of renewable and sustainable energy sources (94.6%), implementing fines and raising their value to reduce the burning of rice straw and infringement on the agricultural environment (91.6%), activating environmental laws to reduce pollution (90%), activating the role of civil society organizations to spread awareness of climate change issues and environmental pollution (87.3%), conducting guidance seminars to raise awareness of the dangers of climate change (85%), cooperation between government agencies, the environment, and civil society organizations (83.6%), enacting laws that criminalize the excessive use of pesticides and chemical fertilizers beyond the recommended levels (82.6%), issue fines and criminalize harvesting crops before the end of the pesticide safety period in order to ensure food safety (80.6%), good urban planning that allows the spread of green areas (78.3%), encouragement organic farming methods to preserve agricultural resources from pollution (75%), and deploying waste collection units in all suburbs and regions to reduce pollution and facilitate recycling (73.3%).

The current study screened the several types of pollutants in the urban environment (pesticides, exhausts, agricultural and domestic waste, industrial activities, and dumping solid wastes). Most of the participants suffered from exposure to pollutants, but they never volunteered to protect the environment. Frequencies and percentages of participants with knowledge of environmental sustainability were high among participants. Frequency and percentages of more sustainable alternative energy sources in the coming decades (sun and water). Most common types of pollution in urban environment were air, water, agricultural, and soil. Four related classifiers were tried to classify the data including, SVM and logistic regression. Participants recommended adopting activities that reduce the presence of pollutants, such as recycling, manufacturing of by-products, and disposing of household waste safely.

Conflict of interest statement

The authors declare that they have no conflict of interest.

Funding statement

This manuscript received no external funding.

Disclaimer

Atef M. K. Nassar, as the Editor-in-Chief of this journal, didn’t participate in the peer review, editorial handling, or decision-making of this manuscript. Full responsibility for the editorial process was under the supervision of another Editor.