Type: Article
Intelligent data analysis for selecting suitable marine sites for mariculture: The use of Aidos intelligent system in Saudi Arabia as a case study
1. Department of Zoology, King Saud University, Riyadh, Saudi Arabia.
2. Cherkassy Branch of Private Higher Education Establishment “European University”, Cherkassy, Ukraine.
Abstract: A severe shortage of fresh water, high evaporation, minimal annual rainfall, absence of rivers, and extreme heat force farmers in Saudi Arabian to cultivate fish, shellfish, shrimp, and other seafood in sea cages. The long coastline along the Arabian Peninsula, with favorable hydrological conditions that ensure stable sea temperatures and oxygen levels, as well as good water exchange, facilitates the implementation of mariculture projects at a lower cost. To ensure high productivity, a suitable marine site for mariculture must be selected. However, many factors are numerical and linguistic and could be vague and fuzzy, making it difficult to formalize the solution. This study aimed to apply automated systems-cognitive analysis to select a suitable marine site that will ensure high mariculture productivity. To achieve this goal, ten mathematical models were synthesized in the Aidos intelligent system, among which, the INF7 model demonstrated the highest identification reliability, and their validity was selected. Of the 120 marine areas included in the training set, 46 were deemed suitable for mariculture, 32 were deemed satisfactory, 25 were deemed poor, and 17 were deemed unacceptable. The results of this study can be used worldwide, as the Aidos system is freely available online, and the user interface can be customized to the most commonly used language.
Keywords: Aidos system, automated system-cognitive analysis, selection of aquaculture sites, mariculture, sustainability
Article Info.
Submitted: 01-20-2025; Revised: 02-04-2026; Accepted: 02-06-2026; Online: 02-08-2026
Cite as: Sanni, M.O, and Ryabtsev, V. (2026) Intelligent data analysis for selecting suitable marine sites for mariculture: The use of Aidos intelligent system in Saudi Arabia as a case study. Animal Reports, 2 (1): 35-58. https://doi.org/10.64636/ar.45
© 2026 by Author(s). This work is licensed under CC BY 4.0.
1 Introduction
The decline of many commercial fisheries, together with the continuous growth in global fish consumption, has stimulated the development of new fish farming technologies to meet increasing public demand (Béné et al., 2015). Mariculture represents a specialized branch of aquaculture in which marine organisms such as fish, shrimp, and oysters are cultivated in seawater. Marine mariculture has developed rapidly, particularly in the form of cage fish farming. The most widespread practice is cage mariculture, which employs deep-sea cages composed of a supporting frame, netting, and anchoring systems. The net forms an enclosed aquatic space that ensures the containment and protection of farmed organisms, including pearl oysters, groupers, starfish, seaweed, and lobsters. To safeguard cages and farmed fish against storm surges and extreme wave action, floating breakwaters are commonly employed.
The Government of the Kingdom of Saudi Arabia actively promotes the development of the aquaculture sector and provides extensive support to fish farmers through various national initiatives. Aquaculture in Saudi Arabia includes the farming of several fish species, such as tilapia, grouper, and sea bream, within controlled environments including ponds, cages, and tanks (Mathew & Alkhamis, 2024). Modern technologies are widely applied to ensure high product quality while minimizing environmental impacts, and producers adhere to strict regulatory frameworks to guarantee food safety and sustainability. As a result, aquaculture makes a substantial contribution to national economic development and provides a reliable source of high-quality protein for the population (Mansour, 2025). Fish farming practices in Saudi Arabia vary depending on farm location, scale, and objectives. Pond systems involve the construction of shallow earthen ponds or artificial lagoons and are particularly suitable for robust species such as tilapia. Cage systems consist of floating structures suspended in the water column and are commonly used for species such as grouper and sea bream, which require stable environmental conditions. Tank-based systems rely on land-based water-filled tanks and allow for intensive production under highly controlled conditions.
When selecting marine areas for cage aquaculture, decision-making is commonly supported by models derived from geographic information system (GIS) data to promote the sustainable and rational use of coastal marine resources (Perez et al., 2005; Windupranata & Mayerle, 2009). A GIS is a system designed to collect, store, analyze, and manage spatially referenced data using a coordinate reference framework, enabling comprehensive analysis of geographic and environmental attributes (Longley et al., 2005; Micael et al., 2015). The selection of suitable mariculture sites requires the collection and analysis of multiple environmental parameters, including water temperature, salinity, pH, visibility, dissolved oxygen, chlorophyll concentration, depth, and current velocity, along with additional auxiliary data (Prema, 2013). Optimal mariculture sites are typically characterized by sufficient depth, relatively low current velocities, and stable current directions. Favorable marine conditions generally include salinity levels of 15–30 % water temperatures of 22–31 °C, and current velocities of approximately 0.65 m s⁻¹, with cage bottoms positioned at least 2 m above the seabed at low tide (Perez et al., 2005; Turner, 2000).
Many high-value marine fish species are carnivorous and traditionally rely on trash fish as feed; however, such feed alone is nutritionally inadequate for optimal growth. Consequently, the availability of high-quality pelleted feed must be considered during aquaculture site selection (Prema, 2013). In offshore cage installations, wind speed and short-interval gust intensity, typically assessed over 15-second periods, are critical design and safety considerations (Turner, 2000). Extreme environmental conditions must also be evaluated when identifying suitable marine locations for fish farms (Chu et al., 2023). GIS-based spatial analyses have been widely applied to generate suitability maps for open-sea maricultural development (Divu et al., 2021; Micael et al., 2015; Nhhala et al., 2022). A satellite image of the Arabian Peninsula, derived from NASA’s Sea-viewing Wide Field-of-view Sensor project, is shown in Fig. 1.
One of the most widely used methods for constructing models to identify suitable mariculture sites is the Analytical Hierarchy Process (AHP), originally developed by Saaty (1980). Despite its popularity, many steps in the AHP rely on manual calculations, which restrict scalability and limit automation. Moreover, standardized environmental factors are multiplied by weighting coefficients assigned subjectively by researchers, making the final results highly dependent on expert judgment (Godinho P. et al., 2011; Stofkova, J. et al., 2022). When performing ASC analysis, the structure and relationships between elements of the system are automatically taken into account, which eliminates the dependence of the generated models on the researcher’s reasoning. Under the leadership of Dr. Francesco Cardia, comprehensive and scientifically grounded recommendations for selecting cage aquaculture sites in Saudi Arabia were developed (Cardia et al., 2017). However, these recommendations do not incorporate mathematical models capable of fully automating the site selection process.
Fig. 1: Satellite image of the Arabian Peninsula (Public Domain).
A comprehensive multi-factor approach has also been applied to evaluate aquaculture site suitability along the northern coast of the Persian Gulf, aiming to identify optimal locations with minimal environmental impact and maximum productivity (Jedari Attari & Farhangmehr, 2025). Factors considered in this assessment included total marine area, average depth, distance from the coastline, sea surface temperature, wind intensity, wave height, current velocity, proximity to protected areas and wildlife habitats, existing fishery activity, and interference from coastal and navigational operations. Marine areas were classified as suitable only if all criteria were satisfied, while areas failing to meet several criteria were deemed conditionally suitable. However, due to the absence of explicit functional relationships between site suitability and influencing factors, such approaches cannot be readily integrated into automated intelligent decision-support systems.
Saudi Arabia has a coastline bordering the Arabian Sea, Red Sea, Sea of Oman, and Persian Gulf. The Red Sea is relatively shallow, with approximately 40 % of its area being less than 100 m deep, making it suitable for mariculture (Zaki, et al., 2025). The maximum depth of the Persian Gulf is 102 m. The average depth is less than 50 m, so it is also well-suited for the placement of marine culture.
Mariculture projects are being widely implemented in countries such as India, China, Japan, Indonesia, Oman, and Saudi Arabia, reflecting the global expansion of marine aquaculture (Nadarajah & Flaaten, 2017). Saudi Arabia has set ambitious targets to increase aquaculture production to approximately 530,000 tons annually by 2030 as part of its national development strategy (Mathew & Alkhamis, 2024). Researchers from King Faisal University and King Abdullah University of Science and Technology have played a significant role in disseminating knowledge and promoting innovative mariculture technologies within the agricultural sector. To ensure long-term food security, it is essential to consider the interconnections among fisheries, mariculture, and agriculture, with particular attention to biodiversity conservation and climate change impacts (Blanchard et al., 2017). Artificial intelligence-based systems provide monitoring and forecasting of the marine ecosystem, and also help reduce costs and minimize negative impacts on the environment (Er-Rousse, Omar et al., 2024).
The objective of this study was to conduct an automated system-cognitive analysis (ASC analysis) of empirical data using the Aidos intelligent system in order to identify the marine territory most suitable for mariculture development. The specific objectives of this study were to:
1. Based on empirical data, compile a training database of cage farm conditions.
2. Generate semantic information models in the Aidos intelligent system, evaluate their adequacy, and select the most reliable model.
3. Conduct research into the modeled subject area, including information semantic analysis of classes and features, cluster analysis of classes, and object recognition features.
Marine area factors for mariculture deployment are measured using various descriptive scales and units of measurement. These data can be high-dimensional, fragmented, noisy, and interdependent. Modeling spaces with factors of varying dimensions is a complex mathematical problem. The modeling object is a marine area, which, depending on the factors acting on it, can be considered unacceptable, poor, satisfactory, or good.
2 Materials and Methods
2.1 Formalization of the subject area
The mathematical framework of automated system-cognitive (ASC) analysis implemented in the Aidos artificial intelligence system is based on systemic fuzzy interval mathematics, enabling the processing of heterogeneous, noisy, and multi-scale empirical datasets (Veniaminovich, 2022). Model construction begins with absolute frequency matrices derived from formalized empirical data, linking factor gradations with class states. Model reliability was evaluated using internal validation metrics. The process of transforming data into information and knowledge is shown in Fig. 2 and is carried out in the order corresponding to the course of the cognitive process: from specific empirical initial data, systemic cognitive models are constructed, and then they move on to examine their stability and, if necessary, to refinement.
Fig. 2. Transforming data into information and its knowledge
The process of constructing mathematical models for automated system-cognitive analysis is based on the calculation of a table of absolute frequencies derived directly from empirically formalized data. In this table, rows correspond to gradations of descriptive scales (factor values), while columns represent classes, that is, the gradations of classification scales. Each cell contains the number of observations reflecting the occurrence of a specific feature value in objects belonging to a particular class. In addition, the transition of a modeled object to a specific future state is identified when the object is influenced by a given factor value. Consequently, establishing a fact requires obtaining information about the object’s characteristics, forming an image of the object based on these characteristics, and subsequently identifying this image by comparing it with generalized images and determining the degree of similarity between them (Veniaminovich, 2012).
The following tasks are solved when performing ASC analysis.
1) Cognitive-target structuring of the subject area.
2) Formalization of the subject area (construction of classification and descriptive scales and gradations and preparation of the training sample).
3) Synthesis of a system of generalized and specific models of the subject area (currently, the Aidos system supports three statistical models and seven system-cognitive models).
4) Assessment of the reliability (verification) of the system of models of the subject area.
5) Increasing the reliability of the system of models, including the adaptation and resynthesis of these models.
6) Solving problems of identification, forecasting, and decision support.
7) Study of the modeling object (process, phenomenon) by examining its models: cluster-constructive analysis of classes and factors; meaningful comparison of factor classes; study of the system of determination of states of the modeled object, non-local neurons, and interpretable direct neural networks; construction of classical cognitive models (cognitive maps); construction of integrated cognitive models (integrated cognitive maps).
The Aidos system favorably differs from known systems in the following parameters:
o It was developed in a universal formulation that does not depend on the subject area; therefore, it is universal and can be applied in many subject areas.
o Provides the transformation of the initial empirical data into information, and it into knowledge and solving problems of classification, decision support, and research of the subject area by studying its system-cognitive model, while generating a very large number of tabular and graphical output forms.
o Provides a stable identification in a comparable form in terms of the strength and direction of cause-and-effect relationships in incomplete noisy interdependent (nonlinear) data of a very large dimension of numerical and non-numerical nature, measured in various types of scales (nominal, ordinal, and numerical) and in various units of measurement;
o It is in full open free access (http://lc.kubagro.ru/Aidos/_Aidos_X.htm), and with up-to-date source texts.
2.2 Data Collection
The following model (S_of_site) was used to select marine areas suitable for mariculture.
S_of_site = <DST, CV, WH, ESS, WT, S_psu, DO, Am, PH_W, N, Ph, BST, DG, DNL, DIA, DTA>,
where: DST - Depth at Spring Tide (m).
CV: Maximum Current Velocity (m/s);
WH - Maximum Wave Height (m);
ESS - Entrainment of Seabed Sediment by Wave (Dean Number);
WT - Water Temperature (°C);
S_psu - Salinity (psu);
DO - Dissolved Oxygen(mg/l);
Am - Ammonia (mg/l);
Ph_W - Water pH;
N: Nitrate (mg/l)
Ph - Phosphate (mg/l);
BST - Bottom sediment type;
DG: Distance to harbor (km)
DNL: Distance to navigation line (km)
DIA: Distance to industrial area (km)
DTA - Distance to tourism area (km).
The ranges of change in the parameters of marine areas intended for the placement of mariculture are given in Table 1 (Windupranata, 2009).
Table 1. Ranges of change in parameters of marine
|
Parameter name |
Parameter ranges |
||
|
small |
optimal |
unacceptable |
|
|
Depth at Spring Tide (m) |
< 5 |
10-50 |
>50-80 |
|
Maximum Current Velocity (cm/s) |
< 5 |
5–30 |
> 50–60 |
|
Maximum Wave Height (m) |
0 |
≤ 1,5–2 |
> 3–4 |
|
Entrainment of Seabed Sediment by Wave (Dean Number) |
2.4 ± 3.2 |
3.2 ± 4.0 |
< 2.4 |
|
Water Temperature (°C) |
5–10 |
15–25 |
> 28–30 |
|
Salinity (psu); |
15–25 |
28–37 |
< 10 > 40 |
|
Dissolved Oxygen(mg/l) |
4–5 |
≥ 6 |
< 3 |
|
Ammonia (mg/l) |
0 |
< 0.02 |
≥ 0.05 |
|
Water pH |
7.5–7.8 |
7.8–8.3 |
< 7.3 |
|
Nitrate (mg/l) |
0 |
0.05 – 1.0 |
> 10 |
|
Phosphate (mg/l) |
0 |
0.05 – 0.1 |
> 0.2 |
|
Bottom sediment type |
Shell rock |
Sand |
Rocky ledges |
|
Distance to harbor (km) |
0.2 |
0.2-0.5 |
< 0.2 |
|
Distance to navigation line (km) |
0.2 |
2-5 |
< 0.2 |
|
Distance to industrial area (km) |
0.5 |
0.5 - 2 |
< 0.5 |
|
Distance to tourism area (km). |
0.5 |
0.5 - 2 |
< 0.5 |
The linguistic variable “Suitability of the marine area” contains the terms: Unacc (Unacceptable). Poor. Fair. Good. The model includes the most important factors influencing the sustainable development of high-productivity mariculture (Veniaminovich, 2016; Perez et al. 2005; Windupranata & Mayerle. 2009). Numerical factors are divided into the following gradations: very small. small. medium. large. and very large. Numerical factors are divided into the following gradations: small. mean. big. A training database was created directly based on the empirical data. The training database included the placement conditions of 120 virtual mariculture sites. Virtual areas are those whose parameters are taken from literary sources (Windupranata, 2009).
2.3 Synthesis and verification of models
This study employed automated system-cognitive analysis using the Aidos intelligent system as its methodological and technological foundation. ASC analysis enables the identification and quantitative evaluation of cause–effect relationships between system components. internal structure. and emergent properties across diverse subject domains. Based on empirical data. formal models were developed to characterize both the magnitude and direction of influence exerted by individual factor values on the future states of the modeled system (Veniaminovich, 2004).
Initial generalization of the training dataset was achieved through the construction of an absolute frequency matrix. in which rows represent influencing factors and columns correspond to target and undesirable future states of the modeled object. On the basis of this matrix, conditional and unconditional percentage distributions were calculated. which subsequently served as the foundation for generating system-cognitive model matrices.
For each dataset, the Aidos system automatically synthesizes ten mathematical models and evaluates their reliability using multiple internal criteria. Model verification included an assessment of first- and second-type errors. namely misclassification and false recognition errors. Reliability metrics applied in this study included Van Rijsbergen’s F-measure and the L1-measure proposed by Veniaminovich (2015). Among the synthesized models, the INF7 model demonstrated the highest identification reliability and was therefore selected for subsequent analyses.
The mathematical models applied within ASC analysis rely on systemic fuzzy interval mathematics. enabling consistent handling of heterogeneous. noisy. and interdependent datasets across different scales of measurement. The central concept underpinning model construction is the estimation of the information content associated with individual factor values and their contribution to transitions between class-defined system states.
The primary form of initial generalization of the training dataset is the absolute frequency matrix (Table 2). In this matrix. rows correspond to influencing factors. columns represent the target and undesirable future states of the studied object. and the cell values indicate the number of observations in which a given i-th factor was present and the object transitioned to a specific j-th state. On this basis. matrices of conditional and unconditional percentage distributions were calculated (Table 3). Subsequently. system-cognitive model matrices were derived from these distributions. as presented in Table 4.
Table 2. Matrix of absolute frequencies (ABS statistical model).
Table 3. Matrix of conditional and unconditional percentage distributions (statistical models PRC1 and PRC2).
Table 4. System cognitive model matrix.
Designations in tables:
i - value of the past parameter;
j - value of the future parameter;
Nij - number of meetings of the j-th value of the future parameter with the i-th value of the past parameter;
M - total number of values of all past parameters;
W - total number of values of all future parameters.
Ni - number of occurrences of the i-th value of the past parameter throughout the sample;
Nj - number of occurrences of the j-th value of the future parameter throughout the sample;
N - number of occurrences of the j-th value of the future parameter with the i-th value of the past parameter throughout the sample.
Iij - a particular knowledge criterion: the amount of knowledge in the fact of observing the i-th value of the past parameter that the object will go into the state corresponding to the j-th value of the future parameter;
Pi - unconditional relative frequency of meeting the i-th value of the past parameter in the training sample;
Pij - the conditional relative frequency of meeting the i-th value of the past parameter at the j-th value of the future parameter.
For each study, the Aidos system automatically generates ten mathematical models and independently evaluates their reliability using multiple criteria. Verification of semantic information models. including the assessment of their reliability or adequacy. can be performed using several approaches implemented within the ASC analysis toolkit of the Aidos system. such as internal validation. external validation. and the bootstrap method. In this study. errors of the first and second kinds—namely. misidentification errors and false recognition errors—were quantified. The resulting criteria used to assess model quality are presented in Fig. 3.
To evaluate the reliability of the models within the Aidos system. Van Rijsbergen’s F-measure and the L1-measure proposed by Veniaminovich (2015) were applied. Fig. 3 illustrates the validity of the system-cognitive models. Among the generated models. The INF3 model demonstrated the highest reliability in object recognition and was therefore selected for further analysis.
For the INF3 model. which is frequently identified as the most reliable. a partial chi-square criterion is employed. This criterion represents the difference between the observed and theoretically expected absolute frequencies. The mathematical models used in ASC analysis within the Aidos system are grounded in systemic fuzzy interval mathematics. enabling comparable processing of large volumes of heterogeneous. noisy. and interdependent data expressed across different measurement scales (nominal, ordinal, numerical) and diverse units (Veniaminovich et al. 2015). The core principle underlying model construction is the calculation of the amount of information contained in a factor value. under the influence of which the modeled object transitions into a specific state corresponding to a given class.
Fig. 3. Model reliability evaluation results.
On the basis of system-cognitive models. the problems of identification (classification, recognition, diagnostics, and forecasting), decision support, as well as the problem of studying the modeled subject area by studying its system-cognitive model are solved. The Aidos system allows you to simultaneously work with three statistical models and seven knowledge models, which allows you to solve the problems of identification, decision making and research of the subject area in all these models according to two integral criteria (Orlov, 2022). ASC analysis and the Aidos system were successfully applied in 8 doctoral and 8 candidate dissertations in economic, Technical, biological, psychological, and medical sciences.
3 Results
3.1 Quality of life recognition results
Recognition is an operation of comparing and determining the degree of similarity of a given specific object with other specific objects or with generalized images of classes. as a result of which a rating of objects or classes is formed in descending order of similarity to a recognizable object. Fig. 4. shows examples of identifying marine areas well suited for mariculture.
An information portrait of a class is a set of data and characteristics collected and analyzed by the Aidos intelligent system. Such a portrait includes information about the strength and direction of the influence of factors on the suitability of a site for placing mariculture. The researcher must first and carefully identify those factors that have the greatest influence on the state of the class. Overall. an information portrait of a class can help the researcher highlight the most important features and reduce the complexity of collecting initial data. Figs 4a and 5b utilize a combined presentation format. combining tabular and graphical formats. This enhances clarity and reduces the length of the research results.
(A)
(B)
Fig. 4. Examples of identifying marine areas for the placement of mariculture: A – Good; B – Unacceptable.
The reliability of identifying marine zones is determined by the similarity module's value. which reflects how closely these zones match the criteria for mariculture suitability. A higher value of the similarity module indicates a more dependable identification process. Model performance was evaluated using internal validation procedures implemented within the Aidos system. including analysis of first- and second-type errors and information-based reliability criteria. In this study, the average similarity module for the zones was 76.5. The higher the similarity modulus value, the more reliable the recognition results. The following range of similarity modulus values for reliable recognition is established: from 12 to 100. If the study yields a similarity modulus less than 12, the results are considered unreliable. An ASC analysis conducted on the training dataset. which comprised empirical data from 120 offshore sites. revealed no instances of false positives or negatives. This finding supports the training dataset's adequacy for assessing data from real offshore locations.
A histogram of the significance of factors determining the suitability of a marine area for mariculture is shown in Fig. 5. The significance of factors in relative units reflects the degree of influence of these factors on the suitability of a marine area for the placement of mariculture. A marine area is well suited for mariculture if the water has very little ammonia, the bottom is sandy. sediment removal is very high, the water acidity is moderate, the wave height is very low, the distance to the tourist area is very large, the amount of nitrates and phosphates in the water is very low, the flow rate is very low. and the water salinity is high.
Fig. 5. Histogram of the importance of factors of a good marine area for the placement of mariculture. Significance values are expressed in relative information units derived from system-cognitive model weights. scaled from 0 to 70. where higher values indicate stronger influence on site suitability.
A histogram of the significance of factors determining the unsuitability of a marine area for the placement of mariculture is shown in Fig. 6. A marine area is unsuitable for the placement of mariculture if the water temperature is very low. The water contains a very large amount of ammonia and phosphates. The removal of bottom sediments by waves is very low, there is very little dissolved oxygen, the wave height is very high, the depth during the spring tide is very small. The distance to the harbor is very small, the salinity of the water is very low, the amount of nitrates is very high, there is mud in the bottom sediments. The distance to the industrial zone and the navigation line is very small, the acidity of the water is very low.
When selecting a site for a mariculture operation, it's important to prioritize those factors that have the greatest impact on the site's suitability. If several factors determine the site's suitability. The remaining factors can be ignored, and the survey can proceed to the next site. This will reduce the intensive labor process of finding the right site for mariculture operation. You can add or remove factors for consideration by adding or removing the appropriate column in the Excel file. You can also change the number of gradations of numerical factors; in this study, the number of such gradations was set to three.
Fig. 6. Histogram of the significance of factors determining the unsuitability of a marine area for mariculture. Significance values are expressed in relative information units derived from system-cognitive model weights (0–70 scale). with higher values indicating stronger contribution to site unsuitability.
The Pareto curve shown in Fig. 7 illustrates the cumulative contribution of descriptive scales to overall classification significance. The most influential 50 % of features account for approximately 72 % of total explanatory power, while the top 30 % contributed roughly 50 %. The monotonic increase of the curve indicates the absence of redundant variables, confirming the efficiency of the selected factor set.
Fig. 7. Pareto curve for the significance of descriptive scales.
To assess the relative suitability of different marine sites, the ASC procedure was applied to two offshore areas with differing environmental characteristics (Table 5). Initially, both areas were provisionally classified as suitable. and ASC analysis was performed. The procedure was then repeated under the assumption that both sites were unsuitable. The resulting similar profiles. presented in Fig. 8 (A and B). demonstrate that the first marine area exhibits a substantially higher correspondence with favorable suitability criteria. indicating superior conditions for cage mariculture deployment. Based on the analysis of the results obtained. We come to the conclusion that the first area of the sea has the best conditions for placing cage mariculture there.
Table 5. Differences in factors between two sea areas.
|
№ Area |
DST |
WH |
ESS |
S_psu |
DO |
Ph_W |
N |
Ph |
DG |
DIA |
DTA |
DNL |
|
Area 1 |
10.25 |
0.37 |
4.26 |
30.5 |
11.75 |
8.35 |
63.00 |
48.25 |
0.77 |
7.7 |
1.51 |
0.51 |
|
Area 2 |
2.8 |
1.98 |
2.1 |
8.8 |
2.9 |
1.99 |
275.00 |
83 |
0.13 |
0.62 |
1.52 |
0.52 |
|
|
|
(A) |
(B) |
Fig. 8. Results of similarity of suitability of sea areas with class gradation: A - Good; B – Bad.
Selecting a marine area without mathematical research will require significant financial and other resources. Furthermore, the experiments will be lengthy. and the results will be unpredictable. Computer modeling will only incur minor costs for the researcher's salary and electricity bill. To determine the suitability or unsuitability of a marine site for mariculture, the following steps must be completed. First, a training set must be developed and fine-tuned, then the study site's factors must be added to the training set. A given site can be assigned to any class graduation. The resulting ASC analysis will confirm or reject the user-specified class gradation's correspondence with the actual value. Discrepancies between the user-specified gradation and the actual value are highlighted in blue. A positive economic impact is expected from using ASC analysis to assess the suitability of a marine area for mariculture. The use of this new innovative methodology will reduce the duration of the study and improve the reliability of the results
3.2 Results of SWOT-analysis
SWOT is an acronym for strengths, weaknesses, opportunities, and threats. SWOT analysis is a widely known and recognized method of strategic planning. However, this does not prevent him from being criticized. often quite fair. justified and reasoned. As a result of a critical review of the SWOT analysis, many of its shortcomings were identified, the source of which is the need to involve experts. in particular. to assess the strength and direction of the influence of factors. It is clear that specialists do this intuitively. based on their professional experience and competence. But the possibilities of experts have their limitations and often for various reasons they cannot and do not want to do this. Thus, there is a problem of conducting SWOT analysis without the involvement of experts. This problem is solved in the Aidos system by automating the functions of experts, that is. measuring the strength and direction of the influence of factors directly on the basis of empirical data. SWOT diagrams of the class of suitability of marine areas for mariculture are presented in Fig. 9. The SWOT diagrams display the 14 most significant relationships. with the connection sign displayed in color (red plus. blue minus). and the value displayed by the thickness of the line. It is possible to display charts with only positive or only negative relationships. When performing a SWOT analysis, only one future target state can be established. but some recommended factors cannot be used due to technological or financial limitations.
|
|
|
(A) |
(B) |
Fig. 9. The SWOT diagrams of the suitability class of marine areas for mariculture: A– Good; B – Unacceptable.
3.3 Cognitive Class Clustering
Clustering is an automatic classification operation in which objects are combined into groups (clusters) such that differences between objects are minimized within groups and maximized between groups. Cluster-constructive class analysis provides the following: calculation of a class similarity matrix; generation of clusters and structures; viewing and printing of clusters. Information about the similarity/difference of classes contained in the similarity matrix is visualized as an agglomerative dendrogram obtained as a result of cognitive clustering. Fig. 10 shows a two-dimensional semantic network of classes. The network nodes correspond to the following qualities of marine areas for mariculture: 1 - satisfactory. 2 - good. 3 - poor. 4 - unsuitable for use. Fig. 10 shows that the two gradations of the class S_of_site Poor and Unacceptable. have a high degree of similarity and therefore can be combined. which will reduce the complexity of the study.
Fig. 10. Semantic 2D class network. Distances between nodes reflect semantic similarity derived from the system-cognitive model. with closer nodes indicating higher similarity between suitability classes.
Fig. 11 shows a two-dimensional semantic diagram that illustrates the quantitative assessments of similarity/difference between classes. It is important to note that this cognitive diagram provides quantitative evaluations of the similarities and differences in factor values derived from a system-cognitive model based directly on empirical data. This approach contrasts with the traditional method. which relies on informal expert assessments grounded in empirical data. experience. intuition. and professional expertise. The 2D cognitive diagrams compare factor values by examining their influence on the modeled object, specifically, its transitions to the states of the relevant classes. thereby indicating how similar or different any two factor values are in terms of meaning. However, the diagram does not specify the exact degree of similarity or difference between the factor values. Although Fig. 11 lacks clarity, it highlights the complexity of the challenge involved in selecting an appropriate offshore site for mariculture.
Fig. 11. Semantic 2D Cognitive Feature Diagram. The diagram visualizes semantic similarity among factor values based on their influence on class transitions. highlighting the multidimensional complexity of mariculture site selection.
3.4 Synthesis of non-local neurons and neural networks
In the Aidos system. non-local neurons are visualized as special graphical forms. which display the strength and direction of the influence of the neuron's receptors on the degree of its activation/inhibition in terms of color and thickness. Fig. 12 shows examples of non-local neurons. In the above fragment of the neural network layer. neurons correspond to factors of the marine area. and receptors to the suitability levels of the mariculture placement. Neurons are arranged from left to right in order of decreasing determination strength. i.e. the results most strictly determined by the factors causing them are on the left. and those less strictly determined are on the right.
|
|
|
(A) |
(B) |
Fig. 12. Nonlocal neurons for of the suitability class of marine areas for mariculture: A – Good; B – Unacceptable. Line thickness and color indicate the strength and direction of factors influence on neuron activation or inhibition.
Fig. 13 shows a fragment of one layer of a nonlocal neural network. In this fragment of a neural network layer. neurons correspond to factors of the marine region. and receptors correspond to the suitability levels of a mariculture site. Neurons are arranged from left to right in order of decreasing strength of determination. i.e. the outcomes most strongly conditioned by the factors that cause them are on the left. while those less strongly conditioned are on the right.
Fig. 13. Non-local neural network.; Neurons correspond to the following gradations of sea areas: 1 – Fair. 2 – Good. 3 – Poor. 4 – Unacceptable; 16 receptors are shown. but in reality, there are 48. Neurons are ordered by decreasing determination strength from left to right. illustrate the relative influence of factors on suitability classification.
4 Discussion
To obtain practical research results, a database must be created in Excel format. and all subsequent operations are performed automatically in the Aidos system using existing applications. This significantly reduces the labor intensity of the study compared to the well-known analytic hierarchy process. The spring 2011 version of the Aidos system supported a training sample size of no more than 100.000 objects; however. in the current version this limitation has been removed. and the system is now capable of processing millions of objects. Nevertheless, there remains a limitation on the dimensionality of knowledge bases. namely. no more than 4.000 classes and 4.000 gradations of factors (Veniaminovich, 2012).
The universal intelligent system Aidos has Eidos is characterized by a highly complex user interface and includes 55 operational modes. excluding the “Exit” mode. As a result, independent learning of the system is extremely difficult. Any operational error may cause the system to halt or may lead to incorrect results. If a user attempts to study several operating modes per day without sufficient explanation. mastering the Aidos system may require several months. To reduce training time, it is recommended to consult the instructional presentations available online, which explain the principles of applying automated system-cognitive analysis to solve applied problems (Al-Ansi et al. 2023).
The productivity of cage aquaculture farms depends not only on marine environmental factors but also on the species being cultivated and their husbandry conditions, including disease control, feeding regimes, stocking density, protection against predation by waterfowl, escape prevention measures, and other management practices (Yavuzcan & Mansour, 2026). When selecting fish species for aquaculture. particular attention should be given to non-native species. While some non-native species adapt successfully to new environments. others may transmit diseases to native species. cause genetic pollution. and contribute to habitat degradation (Zehra et al. 2025).
The key environmental and spatial parameters identified by the ASC-based Aidos analysis are broadly consistent with findings reported in previous mariculture site selection studies. Factors such as water depth. wave height. and current velocity emerged as dominant determinants of suitability. in agreement with earlier GIS-based and empirical assessments emphasizing the need for moderate depths. low wave exposure. and controlled hydrodynamic conditions to ensure cage stability and efficient water exchange (Perez et al. 2005; Turner, 2000; Windupranata & Mayerle, 2009). The strong influence of water quality parameters. including dissolved oxygen. ammonia. nitrate. and phosphate concentrations corresponds with established aquaculture guidelines indicating that elevated nutrient loads and low oxygen levels increase the risk of eutrophication. disease outbreaks. and reduced productivity (Prema, 2013; Divu et al. 2021). Salinity and temperature conditions identified as favorable in this study are also consistent with regional assessments in the Red Sea and Persian Gulf. where relatively stable salinity regimes and warm-water conditions support the cultivation of high-value marine species (Cardia et al., 2017; Mathew & Alkhamis, 2024).
The significance of spatial constraints such as proximity to harbors. navigation routes. industrial zones. and tourism areas aligns with multi-criteria evaluations highlighting the importance of minimizing user conflicts while maintaining operational accessibility (Micael et al., 2015; Jedari Attari & Farhangmehr, 2025). Collectively. these comparisons demonstrate that the ASC-based approach captures well-established suitability drivers while offering a methodological advantage over traditional GIS–AHP frameworks by deriving factor importance directly from empirical data. thereby reducing subjectivity and enabling scalability automated decision support for mariculture planning.
5 Conclusion
As a result of the work carried out using the Aidos system. three statistical and seven system-cognitive models were created. These models directly rely on empirical data to determine the suitability of a marine area for mariculture. using various linguistic and numerical factors. The high technological effectiveness and the ability to support decision-making in selecting a marine area for mariculture are confirmed by the ASC analysis performed by the Aidos intelligent system.
Identifying the strength and direction of the influence of the main factors allows for recommendations on creating conditions that. with maximum determinism. will enable the selection of a suitable marine area for mariculture at minimal cost. Of the 120 marine areas included in the training set. 46 were identified as suitable for mariculture. 32 as Fair. 25 as Poor. and 17 as unacceptable. The accuracy of the classification of offshore areas is 100 %. as no false positives or false negatives were obtained when running the model in the "object-class" mode. which are marked in blue in this mode. The results of the study confirm the suitability of the training dataset for testing real marine sites to select the most suitable ones for mariculture deployment. Future research may extend the application of ASC analysis and the Aidos system to other coastal regions with differing environmental conditions. as well as integrate temporal variability and climate-related factors. The approach may also be adapted for species-specific suitability assessments and combined with GIS-based visualization tools for enhanced decision support.
Acknowledgements
The authors are grateful to Professor Lutsenko E.V. for the opportunity to work in the system Aidos. The authors are particularly grateful to Afeez Mutiat Omobolanle in Nigeria for helping in proofreading and extensive editing of the manuscript.
Author contributions:
Musafau Oloyede Sanni: Data preparation, Investigation, Validation, Writing-Initial draft; Vladimir Ryabtsev: Conceptualization, Methodology, Project management-Project administration, Writing-Review and editing. All authors have read and agreed to the published version of the manuscript.
Ethical approval
This study did not involve live animals or human participants. Ethical approval was therefore not required.
Informed consent
Not applicable
Conflict of interest
The authors declare no competing financial or personal interests that could have influenced this study.
Data Availability statement
Data supporting the findings are available from the corresponding author upon reasonable request.
Declaration of using AI technology
The authors declare that they did not use generative AI and AI-assisted technologies in the writing process of this manuscript.
6 References
Al-Ansi, A. M., Ryabtsev, V. & Utkina, T. (2023). The use of large databases for diagnosing human diseases at an early stage. Indonesian Journal of Artificial Intelligence and Data Mining, 6(2): 197–207. https://doi.org/10.24014/ijaidm.v6i2.24525
Al-Mahish, M. & Alsafra, F. (2025). Factors affecting fish production in Saudi Arabia. Sustainability, 17(21): 9805. https://doi.org/10.3390/su17219805
Béné, C., Barange, M., Subasinghe, R., Pinstrup-Andersen, P., Merino, G., Hemre, G.-I. & Williams, M. (2015). Feeding 9 billion by 2050—Putting fish back on the menu. Food Security, 7: 261–274. https://doi.org/10.1007/s12571-015-0427-z
Blanchard, L. Watson, R. A., Fulton, E. A., Cottrell, R. S., Nash, K. L. & Bryndum-Buchholz, A. (2017). Linked sustainability challenges and trade-offs among fisheries. aquaculture. and agriculture. Nature Ecology & Evolution, 1: 1240–1249. https://doi.org/10.1038/s41559-017-0258-8
Cardia, F., Ciattaglia, A. & Corner, R. (2017). Guidelines and criteria on technical and environmental aspects of cage aquaculture site selection in the Kingdom of Saudi Arabia. FAO.
Chu, Y.-I., Wang, C.-M., Zhang, H., Abdussamie, N., Karampour, H., Jeng, D.-S., Baumeister, J. & Aland, P. A. (2023). Offshore fish farms: A review of standards and guidelines for design and analysis. Journal of Marine Science and Engineering, 11. 762: https://doi.org/10.3390/jmse11040762
Divu, D. N. , Mojjada, S. K., Pokkathappada, A. A. et al. (2021). Decision-making framework for identifying suitable mariculture sites along the northeast coast of the Arabian Sea. Journal of Cleaner Production, 284: 124760. https://doi.org/10.1016/j.jclepro.2020.124760
Er-Rousse, O., & Qafas, A. (2024). Artificial intelligence for the optimization of marine aquaculture. E3S Web of Conferences. 477. https://doi.org/10.1051/e3sconf/202447700102
Godinho, P., Costa, J., Fialho, J., & Ricardo, A. (2011). Some issues about the application of the analytic hierarchy process to R&D project selection. Global Business and Economics Review 13: 26-41. https://doi.org/10.154/GBER.2011.039188
Jedari Attari, M. & Farhangmehr, A. (2025). Assessment of site suitability and capacity for sustainable aquaculture development along the northern Persian Gulf coast. Journal of Human Sciences. https://doi.org/10.22055/jhs.2025.50100.1358
Kukkurainen, Paavo (2017). Fuzzy Logic and Zadeh Algebra. Advances in Pure Mathematics. 07. 350-365. https://doi.org/10.4236/apm.2017.77022
Longley, P. A. Goodchild. M. F. Maguire. D. J. & Rhind. D. W. (2005). Geographic information systems and science (2nd ed.). John Wiley & Sons.
Mansour, Abdallah Tageldein. 2025. Challenges and Innovative Solutions in Sustainable Aquaculture: How Can It Contribute to Food Security and Environmental Protection. Animal Reports, 1 (1): 1-13. https://doi.org/10.64636/ar.8.
Mathew, R. T. & Alkhamis. Y. A. (2024). Contribution of sustainable fisheries and aquaculture to food security in Saudi Arabia. In A. E. Ahmed. J. M. Al-Khayri. & A. A. Elbushra (Eds.). Food and nutrition security in the Kingdom of Saudi Arabia (Vol. 1). Springer. https://doi.org/10.1007/978-3-031-46716-5_8
Nadarajah, S. & Flaaten, O. (2017). Global aquaculture growth and institutional quality. Marine Policy, 84: 142–151. https://doi.org/10.1016/j.marpol.2017.07.018
Nhhala, I., Bahida, A., Nhhala, H., Chadli, H., Baldan, D., Wahbi, M. & Erraioui, H. (2022). Site selection for fish farming using integrated GIS-spatial multi-criteria evaluation and carrying capacity approaches: case study of M'diq bay. Morocco. Egyptian Journal of Aquatic Biology & Fisheries, 26(4). https://doi.org/10.21608/ejabf.2022.254517
Orlov, A. I. & Lutsenko, E. V. (2022). Analysis of data. information. and knowledge in systemic fuzzy interval mathematics. MSc Thesis. Kuban State Agrarian University, Russian. https://doi.org/10.13140/RG.2.2.22539.57120
Perez, O. M., Telfer, T. C. & Ross, L. G. (2005). Geographical information systems‐based models for offshore floating marine fish cage aquaculture site selection in Tenerife. Canary Islands. Aquaculture Research, 36(10): 946-961. https://doi.org/10.1111/j.1365-2109.2005.01282.x
Prema, D. (2013). Site selection and water quality in mariculture. In Training manual on cage culture of marine finfish and shellfish in open sea (pp. 33–43). CMFRI.
Stofkova, J. Krejnus, M. , Stofkova, K. R. , Malega, P. & Binasova, V. (2022). Use of the analytic hierarchy process in sustainable decision-making. Sustainability, 14: 11546. https://doi.org/10.3390/su141811546
Turner, R. (2000). Offshore mariculture: Site evaluation. In J. Muir & B. Basurco (Eds.). Mediterranean offshore mariculture (pp. 141–157). CIHEAM.
Veniaminovich, Lutsenko Evgeny (2004). Intelligent information systems. Textbook for the students of the specialty “Applied Informatics (according to branches)”. Study guide. Kuban State Agrarian University. Russian.
Veniaminovich, Lutsenko Evgeny (2012). Universal cognitive analytical system “Aidos”. Scientific Journal of Kuban State Agrarian University, No. 83 (09).
Veniaminovich, Lutsenko Evgeny (2015). Information and automated control systems. Kuban State Agrarian University. https://doi.org/10.13140/RG.2.2.11978.44480
Veniaminovich, Lutsenko Evgeny (2016). Synthesis of adaptive intelligent measuring systems using ASC analysis and the Aidos system. Scientific Journal of the Kuban State Agrarian University, 116: 1–60.
Veniaminovich, Lutsenko Evgeny (2022). Automated system-cognitive analysis of agrophysical soil indicators. https://doi.org/10.13140/RG.2.2.32110.69446
Windupranata, W. & Mayerle, R. (2009). Decision Support System for Selection of Suitable Mariculture Site in the Western Part of Java Sea. Indonesia. ITB Journal of Engineering Science, 41(1): 77–96.
Zaki, A. , Bashir, B. , Alsalman, A. et al. Evaluating the Accuracy of Global Bathymetric Models in the Red Sea Using Shipborne Bathymetry. J Indian Soc Remote Sens, 53: 277–291 (2025). https://doi.org/10.1007/s12524-024-01981-4
Zehra, S. , Pulukkayi, P. , Boopathi, M. et al. (2025). Introduction of non-native fish species in Red Sea aquaculture: Implications for marine ecosystem integrity. Diversity, 17(4). 296. https://doi.org/10.3390/d17040296
Yavuzcan, Hijran, & Abdallah Tageldein Mansour. 2026. “Technological Innovations Are Driving Precision Aquaculture towards a Future of Transparent, Automated, and Sustainably Efficient Production”. Animal Reports, 2 (1): 16-34. https://doi.org/10.64636/ar.37.