A Comprehensive Sustainable Performance Assessment in Morocco’s Mining Sector Using Artificial Neural Networks and the Fuzzy Analytic Network Process ^†

Abstract

This article provides an in-depth evaluation of sustainability performance within the mining sector by employing the Fuzzy Analytic Network Process (FANP). The assessment centers on five fundamental dimensions: economic, social, environmental, operational, and stakeholder-related factors. FANP facilitates a comprehensive prioritization of both these broad categories and their associated sub-criteria, enabling a well-structured and balanced appraisal of sustainable performance. The methodology is further strengthened by integrating machine learning techniques, specifically a multilayer perceptron, which improves the accuracy and reliability of the multidimensional performance evaluation. Although the study concentrates on the mining industry in Morocco, the developed model is flexible and can be adapted to various other industries and research fields. By filling a significant gap in holistic sustainability assessment, this work offers valuable practical insights to support enhanced management practices and contributes meaningfully to the advancement of sustainable development goals. The findings and approach presented are pertinent to both industry professionals and the academic community alike.

1. Introduction

The mining sector is a key contributor to economic growth but simultaneously faces considerable environmental and social challenges [1]. In Morocco, assessing sustainability performance is vital for promoting responsible and long-lasting mining operations. Although sustainable development has attracted growing attention within the country’s mining industry, existing research often lacks a holistic analysis that encompasses five essential dimensions: economic, environmental, social, operational, and stakeholder-related factors [2]. This study addresses this shortfall by applying the Fuzzy Analytic Network Process (FANP) to evaluate the sustainability performance of a Moroccan mining company. FANP allows for an in-depth analysis of the interactions between these dimensions and ranks them according to their importance, resulting in a more well-rounded assessment. Additionally, the research introduces a versatile model enhanced by a multilayer perceptron (MLP) machine learning algorithm, broadening its applicability to different sectors [3]. This combined approach provides richer insights into sustainability issues and supports more effective, strategic decision-making in the mining industry.

2. Literature Review

The mining sector is a cornerstone of Morocco’s economy, significantly contributing through activities like mineral extraction, phosphate processing, and precious metals mining. To address the sustainability challenges inherent in this industry, a novel approach combines Artificial Neural Networks (ANN) with the Fuzzy Analytic Network Process (FANP). This hybrid methodology demonstrates a strong commitment to sustainable development by leveraging the strengths of both techniques. While ANN provides powerful, data-driven analysis, FANP effectively handles complex decision-making involving multiple criteria, together creating a comprehensive framework for sustainability assessment [3].

Research indicates that in the Moroccan mining context, where economic, environmental, and social factors are closely intertwined, this integrated approach offers valuable insights and practical guidance. It promotes responsible mining operations and helps ensure the sector’s sustainability over the long term. Consequently, the combined ANN and FANP model is gaining recognition as a vital tool for evaluating sustainability performance, with potential applications extending to various other industries worldwide [4].

3. Methodology

This section presents the methodological framework employed in this study, focusing on a thorough evaluation of the mining sector’s multidimensional performance. The core analytical tool utilized is the Fuzzy Analytic Network Process (FANP), which facilitates a structured and nuanced assessment of sustainability criteria across multiple dimensions. The evaluation was implemented using the SuperDecision software (Version 3.2), providing a platform for detailed pairwise comparisons and systematic analysis across economic, environmental, social, operational, and stakeholder-related factors.

To ensure the robustness of the findings before extrapolating broader conclusions, the study incorporated an intermediate validation step. This step involved the direct application of the minimal condition algorithm, which served to consolidate and verify the dimensional and overall performance scores. Acting as a reliability checkpoint, the algorithm helped confirm the consistency and validity of the results, thereby strengthening the foundation for subsequent generalizations.

Building on this validated framework, the methodology was further enhanced by integrating a multilayer perceptron (MLP) machine learning model. The choice of MLP was motivated by its superior capacity to address complex, nonlinear relationships among variables, which traditional methods might overlook. This extension significantly broadened the applicability of the approach, allowing it to adapt to diverse problem domains beyond the mining sector.

Figure 1 visually encapsulates the methodological progression followed in the research, starting with the FANP evaluation, moving through the minimal condition algorithm for validation, and culminating with the machine learning phase using the MLP. This tiered methodology ensures a comprehensive and rigorous analysis of sustainable performance, with the diagram providing a clear overview of the research process and framework [5,6].

3.1. Initiation of the Study

This research focuses on applying the Fuzzy Analytic Network Process (FANP) to evaluate the performance of one of the leading multinational companies operating in Morocco’s mining industry [7,8]. The assessment was conducted using expert judgments gathered from a carefully selected panel of professionals with extensive knowledge of both the specific company and the broader mining sector. These experts’ insights were instrumental in identifying, evaluating, and prioritizing the key criteria and sub-criteria that influence the company’s sustainability performance, thereby ensuring a robust and well-informed analysis.

By relying on this expert-driven approach, the study benefits from practical experience and specialized domain knowledge, which enhances the credibility and reliability of the evaluation outcomes. The research addresses five fundamental dimensions considered critical to sustainable development: economic, environmental, social, operational, and stakeholder transparency [9]. Each dimension comprises a range of relevant sub-criteria that collectively capture the multifaceted nature of sustainability within the mining context.

The FANP integrates the dimensions detailed in Figure 2 together with their associated indicators presented in Figure 3, thereby enabling the development of a comprehensive and systematic assessment framework. This framework enables a nuanced understanding of the company’s overall sustainable performance, accounting for the complex interrelations among various factors and supporting more effective decision-making.

This framework enables an in-depth assessment of the mining company’s sustainability performance. Its organized design guarantees that every crucial element of sustainable development is carefully analyzed, providing valuable insights that aid effective decision-making and inform strategies for future improvement.

3.2. Means and Methods

3.2.1. FANP-Based Decision-Making

The Fuzzy Analytic Network Process (FANP) is an advanced decision-making framework designed to address multifaceted problems by considering both the dependencies between factors and the incorporation of fuzzy logic. Developed by Thomas L. Saaty in 1980 [10], FANP has been widely utilized within operations research and decision sciences. The approach involves building an analytic network that visually represents the interconnections among criteria, sub-criteria, and alternatives. The importance of each element is assessed through pairwise comparisons [10].

By embedding fuzzy logic into the method, FANP effectively handles uncertainty and vagueness, mimicking human reasoning [11]. This integration allows it to process qualitative and imprecise judgments, particularly when dealing with nonlinear or uncertain data. Ultimately, FANP generates numerical outputs that prioritize factors, evaluate performance, and facilitate informed decision-making.

3.2.2. Weighted Sum and Weighted Average

Weight average and weighted sum are both prevalent mathematical methods utilized for merging different values through the application of individual weights [12]. Both methods use weights on components [13], but they differ in their aims and aggregation processes, causing them to be applied differently depending on the analysis requirements. The weighted sum technique involves multiplying each individual value by its assigned weight, followed by summing all these weighted products [14,15]. This calculation yields the overall combined score reflecting the weighted contributions. The mathematical representation of this method is shown in Equation (1):

Weighted Sum = (Value1 × Weight1) + (Value2 × Weight2) +….+ (ValueN × WeightN)

(1)

The weighted sum produces an aggregate score by combining individual elements according to their assigned weights, reflecting the total contribution of each factor. On the other hand, the weighted average calculates a representative measure that incorporates both the size of the values and their relative importance.

This method is especially beneficial when the goal is to compute an average that accurately represents the differing significance of each component [14]. In practice, the weighted average is obtained by multiplying each value by its respective weight, summing these products, and then dividing by the sum of all weights. This process is mathematically detailed in Equation (1).

Furthermore, the algorithm implemented in this research determines the performance level of each dimension by selecting the minimum value among its constituent fields. A step-by-step outline of this algorithm is provided in Figure 4.

3.2.3. Presentation of the “Minimum Condition Algorithm”

• Evaluation at the Field Level: Each specific field within a dimension receives a rating on a scale from 1 to 9.
• Dimension Scoring: The overall score for each dimension corresponds to the lowest rating among its fields, meaning that a dimension’s performance is limited by its weakest area.
• Comparing Dimensions: Scores across all dimensions are analyzed to reveal differences and identify areas that may need improvement.
• Sustainable Development Indicator (D.sd): This value is determined by selecting the minimum score from the economic, environmental, and social dimensions.

D.sd = Min (economic, environmental, social).

5.

Operational and Stakeholder Indicator (D.os): Calculated as the lowest value between the operational and stakeholder dimensions.

D.os = Min (operational, stakeholder).

6.

Performance Levels for D.sd and D.os: The performance categories for these indicators are assigned based on their minimum scores.

7.

Overall Multidimensional Sustainability Score (D.mp): The final sustainability performance is the lesser of the two key indicators, D.sd and D.os.

D.mp = Min (D.sd, D.os).

This evaluation framework uses the minimum condition algorithm, which assumes that overall success is only achieved if every component meets a minimum standard. Performance is measured on a three-level scale from 1 to 9, with the lowest score among dimensions determining the final rating, as depicted in Figure 5.

3.2.4. Artificial Neural Network (ANN)

Artificial Neural Networks (ANNs) are computational frameworks modeled after the human brain’s neural structure, consisting of interconnected nodes that work collectively to interpret data and solve complex problems. Although the human brain excels in certain cognitive functions, computers surpass it when it comes to fast and accurate numerical computations. Among ANN architectures, multilayer perceptrons (MLPs) stand out for their capability to model intricate, nonlinear patterns by employing multiple layers of interconnected neurons, where each connection carries a weight and each node applies an activation function to process information [16,17].

4. Practical Case

4.1. Application of ANN

For this case study, the Fuzzy Analytic Network Process (FANP) was executed using SuperDecisions software [18,19], a platform tailored to facilitate complex decisions through detailed pairwise evaluations. A team of specialists in sustainable performance assessment identified critical links and interdependencies among various dimensions and their corresponding fields [20]. These connections were then mapped into a visual network model (refer to Figure 6), with five distinct clusters representing the main dimensions and nodes symbolizing the examined fields. This visualization provided an organized framework that enhanced the clarity and depth of the analysis by highlighting how different components interact.

Within FANP evaluations, decision-makers commonly utilize the scale values 1, 3, 5, 7, and 9—outlined in

Table 1. Pairwise evaluation scale.

Importance	Explanation
1	Signifies minimal importance or extremely low preference. This implies that the criterion is deemed insignificantly important or inferior when compared to the majority of other criteria.
3	Represents moderate importance or moderate preference. This suggests that the criterion holds a certain level of significance or a moderately preferred position among the other criteria.
5	Indicates an intermediate level of importance or a neutral preference. This means that the criterion is considered to possess a medium degree of importance or a neutral preference when compared to other criteria.
7	Denotes a high level of importance or a strong preference. This indicates that the criterion is regarded as having substantial importance or a significantly higher preference relative to the other criteria.
9	Reflects an extremely high level of importance or preference. This signifies that the criterion is viewed as critically important or holds an exceptionally high preference when compared to other criteria.
2,4,6,8	These even scores reflect intermediate values between the defined categories above.

—to represent the relative preference or importance of different criteria or options whose results are shown in the Figure 7. Each numerical rating is paired with descriptive terms to help convert uncertain or vague judgments into measurable data. Table 1 summarizes the typical meanings assigned to these values during the pairwise comparison procedure.

• Dimensions and Fields Prioritization

This study applies the Fuzzy Analytic Network Process (FANP) to systematically assign importance scores to various fields, facilitating the ranking and quantification of key decision factors. This prioritization process highlights the most influential criteria among many, preventing equal treatment and ensuring focus on those that matter most. After establishing a network that illustrates how criteria and sub-criteria interconnect, the model defines causal and dependency relationships. Importance weights are then derived through a process of pairwise comparisons, where each pair of criteria are evaluated to determine their relative significance. These weights form the basis for calculating the overall importance of each criterion within the sustainability assessment framework, as represented in Figure 8.

b

Sustainable Performance Calculation

To evaluate sustainable performance, weighted values are employed to combine the measurements of criteria and sub-criteria within each dimension. This combination can be achieved through methods like the weighted sum, but the weighted average is often preferred for its ability to produce a single consolidated performance score by dividing the total weighted sum by the sum of the weights. This technique offers a thorough assessment by synthesizing the individual performance scores across different criteria. By applying this method, aggregated weighted values were derived, which accurately represent the relative performance of the various fields under evaluation. Integrating each field’s assigned weight with its performance measurement allows for a detailed appraisal of its overall impact on sustainability.

4.2. Minimal Condition Algorithm and Artificial Neural Network

Before proceeding with further evaluations, the weighted scores were adjusted by applying conventional rounding methods. Subsequently, the minimum condition algorithm was utilized to assess the performance levels of composite indicators—namely D.sd, D.os, and D.mp—by identifying the lowest score among the constituent dimensions. This approach allowed for precise quantification of both D.sd and D.os, which then facilitated the computation of the overall multidimensional sustainable performance metric, D.mp. Building on this foundation, a machine learning technique, specifically a multilayer perceptron, was deployed to analyze a comprehensive dataset exceeding 200,000 records related to the mining company’s performance.

4.3. Outcomes of the Contribution

This study makes two significant contributions. The first is the development of a weighting method that ranks sustainability dimensions based on their impact on overall performance, with particular emphasis on the environment’s pivotal role. The second contribution involves the creation of a predictive model using an Artificial Neural Network (ANN), which demonstrates a high level of accuracy at 94% and a low Root Mean Square Error (RMSE), reflecting its robust forecasting ability. By incorporating weighted inputs, the ANN model effectively identifies complex relationships between dimensions, fields, and performance metrics, resulting in reliable predictions. Together, these approaches form a novel and integrated framework for assessing and forecasting multidimensional sustainability performance, supporting more informed decision-making and promoting sustainability goals.

5. Discussion of Results

Determining the priority weights for different dimensions in assessing sustainable performance within a Moroccan mining firm was a crucial component of the study. This weighting approach enabled the classification of both dimensions and specific fields based on their relative importance, offering essential insights to aid strategic decision-making and direct improvement efforts. By assigning these weights, stakeholders gained a clearer perspective on how each dimension influences overall sustainability, allowing for more targeted allocation of resources to the most impactful areas. The evaluation also demonstrated that the predictive model is dependable and effective, accurately capturing the complexities of multidimensional performance. Additionally, the model’s adaptable nature means it can be customized for other industries by integrating sector-specific indicators like waste management and energy use. This versatile framework serves as a valuable resource for researchers and practitioners alike, supporting comprehensive performance comparisons and promoting more precise management of sustainable initiatives across various contexts.

6. Conclusions

Morocco’s mining sector plays a vital role in both the national economy and the global mineral supply chain, with sustainability becoming an increasingly strategic priority. Supported by government initiatives and infrastructure investments, the country continues to assert its leadership in the global mining landscape. This study employs the Fuzzy Analytic Network Process (FANP) to prioritize five key dimensions of sustainable development: environmental (29%), social (25%), economic (24%), operational, and stakeholder-related aspects (11%). By identifying critical subcategories within each dimension, the research provides stakeholders with actionable insights to guide sustainability efforts. The weighted prioritization underscores the relative significance of each area. Furthermore, the development of a predictive model offers a practical tool for assessing overall sustainability performance, supporting data-driven decision-making. The study delivers a clear and adaptable framework to advance sustainable practices in Morocco’s mining industry, contributing to long-term environmental, social, and economic resilience.