Wind farm sites selection using a machine learning approach and geographical information systems in Türkiye

Abstract

This research highlights the importance of integrating machine learning algorithms with Geographical Information Systems (GIS) applications in the field of renewable energy by finding a suitable site for wind farms due to their importance in preserving the environment to achieve efficiency and cost-effectiveness and reduce the environmental impact of fossil fuel energy sources. Using GIS various factors affecting wind energy localization were processed and analyzed including natural, socio-economic and environmental criteria. Ensemble learning of four supervised machine learning algorithms (Random Forest, K-Nearest Neighbor, Support Vector Machines, Naive Bayes) was used to classify suitable and unsuitable data representing geo-referenced points on the ground with three criteria for each site (wind speed, elevation and slope). The results of the algorithms varied in terms of accuracy and variance, then the results were collected, and the intersection between them was found so that the location classification would be agreed upon in the results of the algorithms used. The aim of using this technique is to reduce the error, increase the accuracy and avoid the bias or variance present in individual models. Accuracy of the algorithms result was respectively (K-Nearest Neighbor, Random Forest, Support Vector Machines, Naive Bayes) (93.022%, 93.018%, 95.095%, 89.553%). The final result is a map using GIS showing the suitable and unsuitable sites of wind farms in the study area (Türkiye) has been chosen as a study area in the research due to several factors that make it suitable for wind energy projects, including its geographical location, which gives it great climatic and terrain diversity, as it is surrounded by seas (Black Sea, Aegean Sea, and Mediterranean Sea), which leads to the activity of seasonal and continuous winds, which contributes to the activity of seasonal and permanent winds. Its drive to develop investment in renewable energy due to economic and population growth has increased the demand for energy and consequently the development of renewable and sustainable energy sources. This research contributes to supporting the global transition to sustainable energy by providing a new methodology for integrating multiple technologies to support a sustainable energy future.

Full text

Vol.:(0123456789)
Discover Computing (2025) 28:22 | https://doi.org/10.1007/s10791-025-09511-7
Discover Computing
Research
Wind farm sites selection using amachine learning approach
andgeographical information systems inTürkiye
OrasFadhilKhalaf1· OsmanNuriUçan2· NaseemAdnanAlsamarai3
Received: 8 November 2024 / Accepted: 6 March 2025
© The Author(s) 2025 OPEN
Abstract
This research highlights the importance of integrating machine learning algorithms with Geographical Information
Systems (GIS) applications in the eld of renewable energy by nding a suitable site for wind farms due to their impor-
tance in preserving the environment to achieve eciency and cost-eectiveness and reduce the environmental impact
of fossil fuel energy sources. Using GIS various factors aecting wind energy localization were processed and analyzed
including natural, socio-economic and environmental criteria. Ensemble learning of four supervised machine learning
algorithms (Random Forest, K-Nearest Neighbor, Support Vector Machines, Naive Bayes) was used to classify suitable and
unsuitable data representing geo-referenced points on the ground with three criteria for each site (wind speed, eleva-
tion and slope). The results of the algorithms varied in terms of accuracy and variance, then the results were collected,
and the intersection between them was found so that the location classication would be agreed upon in the results of
the algorithms used. The aim of using this technique is to reduce the error, increase the accuracy and avoid the bias or
variance present in individual models. Accuracy of the algorithms result was respectively (K-Nearest Neighbor, Random
Forest, Support Vector Machines, Naive Bayes) (93.022%, 93.018%, 95.095%, 89.553%). The nal result is a map using
GIS showing the suitable and unsuitable sites of wind farms in the study area (Türkiye) has been chosen as a study area
in the research due to several factors that make it suitable for wind energy projects, including its geographical location,
which gives it great climatic and terrain diversity, as it is surrounded by seas (Black Sea, Aegean Sea, and Mediterranean
Sea), which leads to the activity of seasonal and continuous winds, which contributes to the activity of seasonal and
permanent winds. Its drive to develop investment in renewable energy due to economic and population growth has
increased the demand for energy and consequently the development of renewable and sustainable energy sources.
This research contributes to supporting the global transition to sustainable energy by providing a new methodology
for integrating multiple technologies to support a sustainable energy future.
Keywords GIS· Machine learning· Wind farm· Machine learning· Sustainable energy
1 Introduction
Addressing the intertwined issues of global energy consumption and environmental impact is indeed critical for sustain-
able development. Implementing reforms to transition towards cleaner and more ecient energy sources is paramount.
This involves not only reducing reliance on fossil fuels but also embracing renewable energy technologies such as solar,
wind, and hydro power [1]. Wind energy is considered one of the most important sustainable energy sources due to
* Oras Fadhil Khalaf, oras.fadil@uosamarra.edu.iq | 1University ofSamarra, Samarra, Iraq. 2School ofEngineering andNatural Sciences,
Electrical andElectronics Engineering, Altınbaş University, Istanbul, Türkiye. 3Al Imam Al Aadum University College, Baghdad, Iraq.
Content courtesy of Springer Nature, terms of use apply. Rights reserved.

Vol:.(1234567890)
Research
Discover Computing (2025) 28:22 | https://doi.org/10.1007/s10791-025-09511-7
the availability of suitable wind resources, environmental conservation and low cost, which gives it a great opportunity
for expansion, growth and development [2]. The world’s total installed renewable energy capacity and its share of the
electricity grid have increased over the past two decades. Today, renewables provide around 40% of the world’s power.
In 2023, the world will add around 510 GW of renewable energy, an increase of almost 50% from the previous year.
This growth rate is the fastest in the past two decades [3]. According to the International Energy Agency (IEA), wind
farms have become widespread in modern energy networks in most countries that support the transition to renewable
energy, and the total global renewable energy capacity is expected to reach 10,800 GW in 2040 [4] Fig.1. This requires
a careful analysis to choose the suitable sites for wind farms and study the factors aecting them to obtain the highest
productivity [5]. Türkiye is one of the countries that seek to invest heavily in the eld of sustainable energy, and its wind
energy production increased from 1375.80MW in 2010 to 11,101.82MW by January 2022, indicating signicant growth,
especially after adding 1797MW between 2020 and 2021, which made Türkiye take its place in the global renewable
energy ranking. Among the eorts and incentives provided by the Turkish government aimed at increasing investment
in this eld are land allocation, tax exemption and price guarantees [6].
Türkiye’s increasing growth in several areas has led to an increasing demand for electrical energy, which is expected
to continue in the future [7]. Selecting suitable sites for wind farms requires a comprehensive assessment of all variables,
criteria, technologies used, and methods, which supports decision-makers in making the right decision in selecting
sites [8]. One of the methods and techniques is integration machine learning with geographic information systems to
analyze spatial data and extract data from it [9]. The integration provides accurate results and improves decision-making
process in several elds, such as urban planning, monitoring environmental changes, agriculture, and natural disaster
management. Integrating machine learning algorithms with geographic information systems can be used to classify of
vegetation, urban areas, and water bodies [10]. Regression analysis and pattern prediction are missing data in spatial
data, such as in predicting air pollution levels in an area [11]. Object detection and feature extraction from remote sensing
data and satellite images such as identifying buildings, roads and vehicles [12]. Spatial clustering and pattern recogni-
tion in spatial data lead to clustering, variance and bias discovery, such as scatter analysis of urban growth [13]. Spatial
analysis and decision support by providing spatial data and information to improve planning, resource allocation, and
risk assessment, considering spatial constraints and patterns [14]. Machine learning and geographic information systems
provide valuable spatial data that can be relied upon to make more accurate and ecient decisions to solve complex
spatial problems. A comprehensive analysis of selection of suitable sites for wind farms must consider economic, envi-
ronmental and social aspects with many inuential factors [15]. The potential of machine learning enhances the achieve-
ment of sustainable development goals by improving renewable energy and the environment [16, 17]. Combining the
fuzzy analytical hierarchy process (AHP) and the demand preference technique based on similarity to the ideal solution
(F-TOPSIS) created a hybrid decision support system to evaluate potential sites for wind power plants. The evaluation
takes into account factors such as wind speed and direction, land use, and distance to power lines [18]. Machine learning
techniques were used to optimize wind power plant site selection in the Philippines, and the analyzed was done using
Support Vector Machines (SVM) [19]. Key tools include machine learning algorithms, datasets, and methods for criteria
weighting and analysis. The study highlights integrating the PROMETHEE II method with GIS for suitability analysis, fol-
lowed by machine learning regression algorithms to generalize suitability indexes across the entire area, introduced a
spatial hybrid decision-making framework for wind farm site selection in northeastern Greece [20]. Split the factors into
two separate groups: economic and environmental. A Spatial Decision Support System (SDSS) is also used to show on a
map an environmental layer solution, an economic layer solution, and a mixed suitability layer on a map. [21]. This study
utilized Rational Basis Function Neural Networks (RBFNN) to transparently assess criteria for wind turbine (WT) installa-
tion locations, integrating results with protected areas and the Land Fragmentation Index (LFI) for identifying socially
Fig. 1 Global electricity gen-
eration mix to 2040 [6]
Content courtesy of Springer Nature, terms of use apply. Rights reserved.

Vol.:(0123456789)
Discover Computing (2025) 28:22 | https://doi.org/10.1007/s10791-025-09511-7
Research
acceptable and energy-ecient sites [22]. GIS and the Analytical Hierarchy Process (AHP) were used to do a wind farm
site t analysis. The AHP was used to give each criterion the right amount of weight based on how important it was, and
the GIS was used to do a geospatial analysis, add many criteria to the proposed index, and make the end suitability map
that showed where the criteria weren’t met [23]. Fuzzy Scaled Weighted Ratio Analysis (F-SWARA) was used to classify
the criteria and determine the priorities in the wind farm installation. In contrast, Fuzzy Measurement Alternatives and
Classication by Leveling Solution (F-MARCOS) were used to determine the most suitable location for the wind farm.
Study area is Sivas province, Central Türkiye [24]. AHP determined the relative weight of each criterion of sites, and GIS
were used to incorporate environmental emissions data from the LCA into the decision-making process. An analytical
model was used to estimate emissions from transportation for each site [25]. The suitable location was selected by
expectation theory and VIKOR multi-criteria ranking method based on determining the criteria weights by combining
Decision-Making and Evaluation Experimental Laboratory (DEMATEL) and Entropy Weight (EW) method; the study area
is China. Previous studies on wind farm site selection have relied on traditional methods, such as multi-criteria decision
analysis (MCDA), A Spatial Decision Support System (SDSS), analytical hierarchy process (AHP), Fuzzy Scaled Weighted
Ratio Analysis (F-SWARA) and geographic information systems (GIS) alone. While these methods are eective, they may
lack the predictive and analytical power of machine learning techniques. Several studies rely on a single machine learn-
ing algorithm, which may lead to biases or inaccuracies and limited integration between machine learning and GIS.
Our study utilizes a combination of machine learning algorithms and GIS tools to provide a comprehensive and reliable
framework for wind farm site selection by combining the outputs of supervised machine learning algorithms (K- Nearest
Neighbor, Random Forest, Support Vector Machines, Naive Bayes) and using GIS to spatially analyze the data and visual-
ize the results to classify sites as suitable or unsuitable for wind farms The methodology provides decision-makers with
a reliable and accurate map of suitable and unsuitable sites for wind farms, supporting Türkiye ’s transition to renewable
energy. The main target of this study is to develop an integrated methodology using machine learning and geographic
information systems (GIS) to identify suitable sites for wind farms in Türkiye. By processing critical factors that aect site
selection, such as wind speed, elevation, slope, socio-economic aspects, and environmental criteria as well as enhancing
decision-making processes, minimizing errors, and improving the accuracy of forecasts, thus contributing to sustainable
energy development and environmental conservation. This research focuses on Türkiye, which has great potential in the
eld of wind energy due to its geographical and climatic diversity.
1.1 Study area
Türkiye is in the Eastern part of the Mediterranean region between 36° N and 42° N latitude and 26° E and 45° E
longitude. Türkiye’s climatic and topographic features have strong contrasts which indicate a characteristic feature
for Türkiye. Türkiye exhibits distinct climatic and topographic variations, which are indicative of its unique nature
[26]. Türkiye is surrounded by three seas (Aegean, Mediterranean, and Black Sea) It has an important impact on wind
speed and distribution, as wind speed is enhanced during the year due to the dynamic interaction between land
and water surfaces due to the difference in temperatures between the land and water surfaces, with winds moving
from sea to land during the day (sea breeze) and from land to sea at night (land breeze). Sea streams also affect the
distribution of temperature and humidity, which supports the formation of suitable sustained winds. Greece and
Bulgaria are situated on one side, while Armenia and Georgia are located on the other side. Iraq and Iran are located
to the southeast, while Syria is situated to the south. Türkiye consists of 81 provinces. The country is divided into seven
distinct regions: Marmara, Aegean, Mediterranean, Southeast Anatolia, East Anatolia, and Central Anatolia. Türkiye
is a prominent nation in the field of sustainable power [27]. Türkiye’s topographic diversity of the Taurus Mountains,
southern and northern Anatolia, the Anatolian plains, the seas, and its location between two continents made it a
source of wind activity throughout the seasons, which allows it to be one of the ideal regions for wind energy [28].
The total area of it is 783,562 Km2 (Fig.2).
Content courtesy of Springer Nature, terms of use apply. Rights reserved.

Vol:.(1234567890)
Research
Discover Computing (2025) 28:22 | https://doi.org/10.1007/s10791-025-09511-7
2 Data set
2.1 Collection andpreprocessing ofdata
This study collected data on topographic, climatic, geological and hydrological factors based on the extensive lit-
erature on wind farm site selection, and different criteria were selected to reflect local conditions. These criteria are
divided into natural, social, economic and environmental Table1. The first section represents the processing and
analysis of the data types collected from satellite images of wind speed, slope and elevation in the form of a Digital
Elevation Model (DEM), and maps of the land use, existing wind farms in Türkiye and converting them into linear
data in the form of points. A dataset and training data were generated to be the inputs for the machine learning
algorithms used in our study. The algorithm’s training data represents the characteristics present in the previously
established wind farm site that are within the required criteria.
2.1.1 Natural factors
Critical factors for site wind turbines are (wind speed, elevation, and slope) Table3. Wind speed is essential for opti-
mal electricity generation. This study uses the Global Wind Atlas to provide monthly wind speed data 100m above
ground. The data assesses wind potential by calculating wind speed distribution curves for each grid cell. Months
with above-average wind speeds are identified, marking the peak season. Wind speed is the most important fac-
tor in choosing a site for a wind farm. A wind speed of more than 3.5m/s is the minimum required to rotate and
Fig. 2 Map of Türkiye
Content courtesy of Springer Nature, terms of use apply. Rights reserved.

Vol.:(0123456789)
Discover Computing (2025) 28:22 | https://doi.org/10.1007/s10791-025-09511-7
Research
Table 1 Details of the data
Data Source Year Description Type
Wind speed Global Wind Atlas 2022 Data contains the wind speed
distribution for each grid cell, with
months with higher than average
wind speeds identied to charac-
terize peak seasons
Shape le (points)
Elevation United States Geological Survey
Earth Explorer 2014 Provides land elevation data at sea
level for each geographic point DEM
Slope United States Geological Survey
Earth Explorer 2014 Represents the degree of inclina-
tion of the earth for each geo-
graphic point
DEM
Socio-economic data Urban areas MapCruzin.com 2014 Determine the distance between
proposed sites and residential
areas
Shape
le(points,
polygon,
lines)
Transportation networks MapCruzin.com 2014 Data contains the location of major
roads and railroads
Power grids National Electricity Grid Maps 2019 Location of power plants and major
transmission lines
Environmental data Nature reserves and bird migration
routes Protected Planet (protectedplanet.
net) 2019 Data contains the location of nature
reserves and migratory bird areas Shape
le(points,
polygon,
lines)
Land use European Environment Agency 2009 Identies types of land use, such as
agricultural and forestry areas
Water bodies European Environment Agency 2009 Location of large lakes and rivers
Content courtesy of Springer Nature, terms of use apply. Rights reserved.

Vol:.(1234567890)
Research
Discover Computing (2025) 28:22 | https://doi.org/10.1007/s10791-025-09511-7
operate a turbine. Sites with stable winds at optimal rates should be chosen to run the turbines economically to
ensure optimal performance. Elevation is a critical factor in wind farm site selection, as it is directly related to wind
speed and turbine efficiency. Regions with higher elevations often have more regular and stronger winds, which
increases the productivity of the turbines. However, higher elevations may also increase construction and transpor-
tation costs due to challenging landforms. Sites under 2500m are considered ideal to balance construction cost
and accessibility. Figure3 shown X axis represents (Wind Speed) Y axis represents (Slope): We note that the slope is
spread between 0 and 70 degrees. However, the high density of points in the enclosed part (0–30 degrees) and the
points are within the suitable slope as well as for points that represent wind speed greater than 3.5 Ms, making it a
positive factor when choosing wind farm locations because low slope makes easy construction and maintenance
and minimizes costs. X axis represents (Wind Speed) Y axis represents (Elevation): There is a clear variation of wind
speed with elevation, as the general trend shows that wind speed increases with increasing elevation. This pattern
is expected because higher elevation have fewer ground obstacles, which allows the wind speed to increase. Few
points can be observed at low elevation and high wind speeds. X axis represents (Slope) Y axis represents (Elevation):
shows the slope does not depend much on elevation, as the points are evenly distributed across different elevation
values. This may indicate that in the studied areas, slope is a relatively independent factor from elevation; the slope
Fig. 3 Distribution of data (wind speed, slope, and elevation)
Table 2 Descriptive statistics
of the data Data Count of data points Minimum value Maximum value Mean
Wind speed 7,204,235 0.42 16.2 5.1
Slope 7,204,235 0 76.6 11.1
Elevation 7,204,235 0 50.7 × 100 13.15 × 100
Content courtesy of Springer Nature, terms of use apply. Rights reserved.

Vol.:(0123456789)
Discover Computing (2025) 28:22 | https://doi.org/10.1007/s10791-025-09511-7
Research
could be a more important criterion in determining suitable sites. Table2 shows descriptive statistics data of factors
(wind speed, elevation, and slope).
2.1.2 Socio‑economic factors
Wind farms should be located at a safe distance from urban areas and residential communities to avoid noise and moving
shadows and ensure the safety of the population, as well as taking into account the future expansion of the population
and the culture of the community in accepting the view of wind turbines; some people see it as visual pollution, and
some people see the rotation of turbines with the wind and landscapes as a beautiful view. The shorter distance between
the wind farm site and the buer zone for transmission networks contributes to solving the problem of transporting
heavy equipment and reducing the time of access to the site, which leads to reduced costs and increases the speed of
operation and maintenance. The greater distance between the wind farm site and airports, the more important it is to
avoid interference with air navigation systems to ensure the safety of air operations and compliance with international
standards. Keeping wind turbines close to the national power grid is critical in reducing transmission costs and minimiz-
ing losses [29].
2.1.3 Environmental factors
Many environmental factors must be considered when selecting wind farm sites from areas of ecological importance and
plant diversity, as wind farms aect vegetation, especially in agricultural and forested areas. Maintaining a buer zone
aims to minimize negative impacts on the environment and maintain a balance between sustainable development and
nature-based development. This will help protect natural resources and avoid conicts with other land uses. Careful land
use planning is necessary to minimize the impact on vegetation and consider agricultural use around wind turbines. It is
necessary to consider the impact on protected areas and bird migration routes. Protected areas are considered habitats
of high ecological importance to ensure the protection of biodiversity and minimize human intervention, as this crite-
rion reects the projects’ commitment to sustainable development and environmental protection. Additionally, water
bodies represent sensitive ecosystems that must be protected from damage or environmental changes and avoid the
impact of wind farm construction on water quality and aquatic life. Wind farms should be located at a safe distance away
from ood-prone areas to protect the turbines. Table3 shows the criteria for selecting sites for wind farms, taking into
consideration natural, social, economic and environmental criteria to exclude unsuitable sites for wind farms by setting
specic criteria for each factor (Fig.4).
3 Machine learning inGIS
Machine learning is a branch of articial intelligence (AI) that deals with developing programs and systems that can learn
and measure from data, extract patterns, and make decisions based on that data. "The ability of the system to correctly
interpret external data, to learn from such data, and to use those learning’s to achieve specic goals and tasks through
exible adaptation" [34]. Machine learning algorithms can build a mathematical model based on a set of data called
"training data" to produce results in the form of predictions or classications without being explicitly programmed to
Table 3 The list values for
criteria and buer zone Criteria Exclusionary criteria Buer zones References
Natural Elevation < 2500m [19, 30, 31]
Slope < 30%
Wind speed > 3.5m
Socio-economic Roads 0–500m [32]
Urban areas 0–2500m
Airports > 3500m
Environmental Land cover/ land use > 500m [33]
Water bodies > 500m
Protected areas 0–2000m
Content courtesy of Springer Nature, terms of use apply. Rights reserved.

Vol:.(1234567890)
Research
Discover Computing (2025) 28:22 | https://doi.org/10.1007/s10791-025-09511-7
do so [34]. One of the types of machine learning algorithm is supervised learning, which adopts a mathematical model
of a dataset with two inputs and desired outputs, which are used for either classication or regression [35]. The other
type is unsupervised learning, which adopts a mathematical model to identify hidden patterns in dataset by learning
from environments and determining data patterns accordingly. This paper used four supervised learning algorithms
(K-Nearest Neighbor, Random Forest, Support Vector Machines, Naive Bayes). Machine learning (ML) is a relatively new
concept compared to Geographic Information Systems (GIS). Machine learning involves a computer program learning
from various experiences to enhance its performance on specic tasks. The program’s measurable performance improves
with increasing experience and insights from executing these tasks. The machine uses assessments and predictions based
on the data it receives [36]. Supervised machine learning algorithms can be trained using pre-labelled data to classify a
dataset or predict future events from unseen data, starting from analyzing the training data set to the learning algorithm
that predicts the output values. The system can provide targets for each new input after sucient training. Furthermore,
the algorithm can compare its output with the correct output and nd errors to adjust the model accordingly. Increas-
ing application of ML techniques in various elds, including GIS. Examples of these algorithms are KNN, Random forest,
Support vector machine (SVM), and Decision tree, etc. [37]. The reason for selecting the techniques used in this study
is the integration of machine learning and GIS, as machine learning provides the ability to analyze and process data
of dierent and large proportions in an eective manner, which leads to the discovery of patterns inside the data and
predicting results based on training data, which increases the accuracy of results and reduces bias and variance through
the use of Ensemble Learning techniques (Ensemble Learning). GIS is an eective tool for collecting, processing, analyz-
ing, and representing spatial data as spatial maps showing suitable and unsuitable sites. Integration with ML enhances
the accuracy of spatial decisions by combining spatial data analysis with the predictions generated by ML algorithms.
4 Methodology
Figure5 explains the methodology used in this study to select suitable wind farm sites in Türkiye. It includes Phase I,
which identies the problem and criteria that dene the specications of the suitable sites and collects data in multiple
forms and sources. Phase II analyzes and processes data using GIS software tools that include maps (Urban areas, trans-
portation routes, reserves, lakes and water bodies, airports) and thus exclude them from being suitable site for wind
farms. After the exclusion of the all land use, a map of the nominated areas for establishing wind farms was created and
converted from raster data to vector data. Each point represents coordinates on the ground and has characteristics (wind
speed, elevation, and slope). Create maps of existing wind farm sites in Türkiye (Fig.6). Each site has the appropriate
characteristics (wind speed, elevation, and slope) and is converted into a formula to train data for learning algorithms.
Fig. 4 Buer zone of all land use
Content courtesy of Springer Nature, terms of use apply. Rights reserved.

Vol.:(0123456789)
Discover Computing (2025) 28:22 | https://doi.org/10.1007/s10791-025-09511-7
Research
Fig. 5 Framework stages
Fig. 6 Wind farm Kürekdagi in Çınarcık region and all existing wind farms in Türkiye
Content courtesy of Springer Nature, terms of use apply. Rights reserved.

Vol:.(1234567890)
Research
Discover Computing (2025) 28:22 | https://doi.org/10.1007/s10791-025-09511-7
Phase III Uses four machine learning algorithms. The same dataset and training data were entered for each individual to
classify suitable and unsuitable points for each algorithm whose results and accuracy were. Phase IV use GIS to represent
algorithm outputs and nd the nal result that represents suitable and unsuitable sites and represent it on the map.
5 Discussions andresults
5.1 K‑Nearest Neighbor (K‑NN) algorithm
The K-Nearest Neighbor (K-NN) algorithm is used in various elds, such as fault diagnosis, power system protection,
and medical detection. Because of its high eciency with minimal misclassify compared to other supervised learning
classiers, it can handle major class problems with comprehensive training data [38]. Advantages (simplicity and ease
of performance, adaptability to multi-class problems) disadvantages (increased computational cost, intensive memory
consumption). Using the data that represents the candidate sites for all the sites for the study area as a dataset to be
the input for the algorithm that works to classify them as suitable or unsuitable sites based on the training data that
represents the currently established wind farms, the result of the classication was as in Fig.7 with accuracy 93.022%.
Figure8 shows the average values of the criteria of wind speed, slope, and elevation for the sites classied as suitable
and unsuitable using the algorithm. In Fig.8a, we can see that the algorithm classied some sites as suitable, even though
their values were outside the criteria dened in the study, indicating that there were exceptions in the classication pro-
cess. Figure8b shows the classication results for the sites that were determined to be unsuitable. It is notable that all
of these sites were within the dened criteria, except for the slope criteria was the decisive factor in categorizing them
as unsuitable. The slope value in these sites exceeded 30%, which is the maximum acceptable value according to the
approved criteria, making them practically unsuitable for wind energy development. The error rate of the classication
Fig. 7 Result of K-NN algorithm classify
Fig. 8 Mean of wind speed, slop, and elevation over wind speed, (KNN)
Content courtesy of Springer Nature, terms of use apply. Rights reserved.

Vol.:(0123456789)
Discover Computing (2025) 28:22 | https://doi.org/10.1007/s10791-025-09511-7
Research
results using the algorithm was 6.978%. This error can be explained by several factors, including the sensitivity of the
algorithm used, the KNN algorithm, to boundary values. Although this algorithm is accurate in many applications, it can
have diculty distinguishing between locations with values that are close to the classication boundary.
5.2 Random Forest algorithm
The Random Forest algorithm is one of the common algorithms in machine learning. It contains many decision trees to
provide accurate results with the lowest error rate [39]. The algorithm creates trees to choose the best classication for the
test data sets and randomly selects data samples to determine the best results [40]. The non-linear method is the method
used by the algorithm to discover relationships between features. This method makes it a powerful tool for classication
and regression modeling. It does not cut trees, unlike other tree-based algorithms. It partitions random subsets of data at
each tree node, increasing the tree set’s diversity and improving performance [41]. Advantages (high accuracy, handles
over tting well) disadvantages (computational complexity, memory usage, poor interpretability). Random Forest is an
eective tool for large-scale and multivariate pattern recognition [42].Using the same training data and data set used in
the previous algorithm, we obtained the results shown in Fig.9 with an accuracy of 93.018%.
Figure10a shows that some sites were classified as suitable even though they do not match the criteria adopted
in our study, especially in slow wind speeds. In Fig.10b some sites were classified as unsuitable even though their
values were within the specified criteria. This normal has an error rate 6.982% for several reasons, one of the most
prominent causes of bias in the results of the algorithm used, as the algorithm shows high sensitivity to values that
fall on the boundaries of the specified criteria when some values are close to these boundaries, such as wind speed,
slope, or elevation, this can lead to fluctuations in the decision-making process between categorizing the site as
“suitable” or “unsuitable.” Additionally, the variance in classification accuracy may be due to the nature of the Random
Fig. 9 Result of Random Forest algorithm classify
Fig. 10 Mean of wind speed, slop, and elevation over wind speed (RF)
Content courtesy of Springer Nature, terms of use apply. Rights reserved.

Vol:.(1234567890)
Research
Discover Computing (2025) 28:22 | https://doi.org/10.1007/s10791-025-09511-7
Forest algorithm, which relies on splitting the data into multiple decision trees. In some cases, this process may lead
to conflicting final decisions when the input values are close to the boundaries between classifications.
5.3 Support vector machines algorithm
Support vector machines (SVM) have been developed in machine learning and applied to several applications, rang-
ing from biological data processing to medical diagnosis, time series prediction to face recognition, geographic data
classification, and other uses [43]. An advantage of a support vector machine is that the model being built is explicitly
based on a subset of data points and support vectors that help interpret the model [44]. Advantages (efficient in
high-dimensional spaces, concentrated to over fitting) disadvantages (computational complexity, sensitive to selec-
tion of seed and hyper parameters). The SVM algorithm used the same dataset and training data used as inputs in
the previous algorithms, and the result is shown in Fig.11 with an accuracy of 95.095%.
Figure12 shows the average values of the parameters wind speed, slope, and elevation for suitable and unsuitable
sites as determined using the SVM algorithm. The results show a clear contrast between suitable and unsuitable sites,
highlighting the efficiency of the algorithm in classifying sites based on the available data. It is worth noting that
these results converge significantly with those obtained using the KNN algorithm, with a low error rate of 4.905%,
reflecting the accuracy and efficiency of both algorithms. The average values for suitable sites indicate ideal condi-
tions for wind power generation, characterized by higher wind speeds, moderate slopes, and favorable heights for
turbine installation. In contrast, unsuitable sites show values that deviate from these optimal limits, enhancing the
reliability of the classification process
Fig. 11 Result of support vector machines algorithm classify
Fig. 12 Mean of wind speed, slop, and elevation over wind speed, (SVM)
Content courtesy of Springer Nature, terms of use apply. Rights reserved.

Vol.:(0123456789)
Discover Computing (2025) 28:22 | https://doi.org/10.1007/s10791-025-09511-7
Research
5.4 Naive Bayes algorithm
The Naive Bayes algorithm calculates a set of probabilities using frequency counting and sets of values in a given
data set. The Bayesian algorithm assumes that all attributes are independent, given the value of the class variable.
The Naive Bayes algorithm performs well and learns quickly across different supervised classification problems [45].
Naive Bayes classification is suitable for general classification predictions. Several successful real-life applications
use the Naive Bayes classification, such as weather forecasting services, customer credit ratings, weather forecast-
ing, health status ratings, etc. [46]. Advantages (simplicity and ease of performance, quick training and prediction)
disadvantages (powerful independence assumption, determined to linear decision borders). The NAB algorithm
used in this study to classify the data based on the same dataset and training data used as inputs in the previous
algorithms, and the result is shown in Fig.13 with an accuracy of 89.553%.
Figure14a shows the results of the algorithm classify, which classified the sites as suitable, and most of the values
were within the standards, except for some of them, where the low-speed wind factor was outside the criteria. Fig-
ure14b shows that some sites were classified as unsuitable, although their values are ideal for suitable use. The reason
is that the Naive Bayes algorithm has a conditional independence technique, which assumes that the features are
independent. However, they are not independent in some cases, and this wrong assume of independence between
factors sometimes leads to inaccurate classifications, especially in cases where the relationships between criteria are
complex and non-linear. For example, a site may be considered optimal based on a combination of factors, but if each
factor is evaluated individually as the Naive Bayes algorithm would evaluate, it may ignore the integrating effect of
these factors, leading to inaccurate results. The error rate for the algorithm classify result is 10.447%.
Fig. 13 Result of Naive Bayes algorithm classies
Fig. 14 Mean of wind speed, slop, and elevation over wind speed, (NB)
Content courtesy of Springer Nature, terms of use apply. Rights reserved.

Vol:.(1234567890)
Research
Discover Computing (2025) 28:22 | https://doi.org/10.1007/s10791-025-09511-7
Table4 shows the model performance analyses for classication evaluation, including parameters (accuracy, error
rate, recall, precision and F-measure) for the four algorithms.
Through the results we obtained from the algorithms (K-Nearest Neighbor, Random Forest, Support Vector Machines,
Naive Bayes) that classied the results data into two categories, suitable and unsuitable, with varying degrees of accuracy
and representation, it was found that the intersection between the results of the four algorithms has a result agreed
upon by the four algorithms used, i.e. the site that was classied as suitable or unsuitable for wind farms is calculated as
the nal result. Figure15 represents a map of the nal sites, suitable in green and unsuitable in red, which were agreed
upon by the four algorithms, and this gave more accurate and reliable results than if a single algorithm was used.
Figure16 represents nal result of the mean values of the criteria (wind speed, elevation, and slope) that represent
suitable site for wind farms, and their values are within the required criteria values Table1. The mean elevation value of
suitable sites does not exceed 2500m (each elevation value in this study is multiplied by 100 in the real world). We note
Table 4 Shows the model
performance analyses Algorithm Accuracy (%) Error rate (%) Recall (%) Precision (%) F-measure (%)
Random Forest 93.018 6.982 96.38 94.35 95.3
K-Nearest Neighbor 93.022 6.978 95.62 95.03 95.33
Support vector machines 95.095 4.905 97.31 95.57 96.43
Naive Bayes 89.553 10.447 91.18 95.6 93.4
Fig. 15 Map of suitable and unsuitable sites resulting from the intersection of the results of the four algorithms
Fig. 16 Mean values of the criteria of suitable sites
Content courtesy of Springer Nature, terms of use apply. Rights reserved.

Vol.:(0123456789)
Discover Computing (2025) 28:22 | https://doi.org/10.1007/s10791-025-09511-7
Research
the mean slope value of suitable sites within the required criteria, and the wind speed is not less than the speed required
to start the turbine rotation close to 3.5m per second and above.
Figure17 represents nal result of the mean values of the criteria (wind speed, elevation, and slope) for unsuitable
sites, as the slope criterion was decisive in classifying the site as unsuitable for wind farms due to the nature of the study
area’s topography, while the wind speed criterion was within the appropriate criterion value in most sites. The elevation
criterion was appropriate in a few sites, but the wind speed and slope criteria were not appropriate, and it is not appro-
priate in most sites. The site must meet the three criteria to be suitable for wind farms.
Figure18 shows that all the sites of previously established wind farms in Türkiye fell within the areas suitable for wind
farms we obtained in our study. This proves the validity and reliability of the study’s results and the method used in it.
The integration of machine learning with geographic information systems (GIS) provides powerful tools for analyzing
spatial data in wind farm location studies with multiple benets. GIS is used in spatial data such as digital maps, analyzing
factors such as wind speed, slope, and elevation, as well as social and environmental factors, to give a visual represen-
tation of the data, but decisions based on this data may be prone to human error. When combined GIS with machine
learning, it more accurately classies and analyzes this data by detecting hidden patterns in the data and reduces error
by using mathematical models to auto-classify sites based on the input data. Machine learning can predict optimal areas
that have yet to be explored based on the available data and the results of the models used, enhancing the sustainability
of the decision-making process.
Fig. 17 Mean values of the criteria of unsuitable sites
Fig. 18 Previously established wind farms in Türkiye
Content courtesy of Springer Nature, terms of use apply. Rights reserved.

Vol:.(1234567890)
Research
Discover Computing (2025) 28:22 | https://doi.org/10.1007/s10791-025-09511-7
6 Conclusion
The study showed the importance of shifting to sustainable energy with its abundant natural resources, including
wind energy in Türkiye, which represents an environmental and economic solution to energy issues, which enhances
the efficiency of natural resource utilization and contributes to the achievement of global sustainable development
goals, as both growing and developing countries can benefit from this methodology in achieving their renewable
energy goals. The study also supports global efforts to reduce carbon emissions by improving the efficiency of wind
farm siting. GIS and its applications are essential in collecting, processing, and analyzing several types of spatial
data to form a dataset or training data for many algorithms in the field of artificial intelligence. This study used the
integration between machine learning algorithms (KNN, SVM, RF, NB) and geographic information systems (GIS) to
enhance classification efficiency and provide easy-to-understand data visualization for decision makers. In addition,
the use of ensemble learning to reduce bias and variance is an innovative step that enhances the reliability of the
results and supports strategic planning. The methodology used in this study using supervised learning algorithms
can be applied to other areas of study. The study showed that the SVM algorithm gave the highest accuracy (95.095%)
due to its advantages in finding the optimal decision boundaries between attributes in a dataset where different
attributes are separated by a large distance. The NB algorithm gave the lowest accuracy (89.553%) due to its condi-
tional independence method, which assumes that the attributes are independent. However, they are sometimes not
independent, leading to inaccurate results when dealing with data with complex correlations between variables. The
dataset and training data resulting from the GIS processing of the study’s raw data can be used and applied using
other AI algorithms with more various parameters depending on the type of study. Although the study’s positive
results are based on a static dataset, it did not consider recent climate changes and disturbances, which reduces its
ability to adapt to future conditions. The study called for additional research to expand the scope of the criteria and
incorporate temporal data and predictive analytics to ensure the sustainability of the results and their long-term
effectiveness. Utilize big data and integrate deep learning to extract more patterns in data through big data process-
ing techniques and satellite data analysis. Optimize the use of unsupervised machine learning to discover hidden
relationships between different factors.
7 Future work
1. Integrating AI technologies with other applications opens up prospects for expanding its application in more areas
like healthcare, nance, transportation, and more.
2. Combining real-time analysis with real-time data enhances prediction and choosing the right decision in regulating
electricity consumption, wind speed predictions, self-maintenance, disaster prediction, etc.
3. Developing cognitive computing innovation models in sustainable energy and environmental conservation. These
models can help simulate scenarios, optimize resource use, and reduce environmental impact. Real-world applications
might include smart grids for energy eciency or AI-based tools for environmental monitoring and conservation.
4. Optimizing multi-criteria decision analysis (MCDA) with ensemble learning: Integrating more progressive techniques
into MCDA can improve decision-making processes, particularly in more complex and multi-critical scenarios such
as site selection. Optimizing ensemble learning methods within MCDA frameworks can lead to more accurate, more
eective, and more informed decision-making
Acknowledgements The authors would like to acknowledge the support of University of Samarra, Samarra, Iraq and Altinbas University,
Istanbul, Turkey for their valuable support.
Author contributions Conceptualization, Oras Fadhil Khalaf.; methodology, Oras Fadhil Khalaf; Software Osman Nuri Ucan.; validation, Naseem
Adnan Alsamarai.; Formal analysis, Oras Fadhil Khalaf.; Writing—original draft preparation, Oras Fadhil Khalaf.
Funding The authors did not receive support from any organization for the submitted work.
Data availability The dataset generated during the study is available and can be shared by the corresponding author upon reasonable request.
Content courtesy of Springer Nature, terms of use apply. Rights reserved.

Vol.:(0123456789)
Discover Computing (2025) 28:22 | https://doi.org/10.1007/s10791-025-09511-7
Research
Declarations
Ethics approval and consent to participate Not applicable.
Consent for publication Not applicable.
Research involving Human Participants and/or Animals Not applicable.
Competing interests The authors declare no competing interests.
Open Access This article is licensed under a Creative Commons Attribution-NonCommercial-NoDerivatives 4.0 International License, which
permits any non-commercial use, sharing, distribution and reproduction in any medium or format, as long as you give appropriate credit to
the original author(s) and the source, provide a link to the Creative Commons licence, and indicate if you modied the licensed material. You
do not have permission under this licence to share adapted material derived from this article or parts of it. The images or other third party
material in this article are included in the article’s Creative Commons licence, unless indicated otherwise in a credit line to the material. If
material is not included in the article’s Creative Commons licence and your intended use is not permitted by statutory regulation or exceeds
the permitted use, you will need to obtain permission directly from the copyright holder. To view a copy of this licence, visit http:// creat iveco
mmons. org/ licen ses/ by- nc- nd/4. 0/.
References
1. Zhao C, Wu Y, Dai X, Han J, Dong B, Huang L-B. Calliopsis structure-based triboelectric nanogenerator for harvesting wind energy and
self-powerd wind speed/direction sensor. Mater Des. 2022;221: 111005. https:// doi. org/ 10. 1016/j. matdes. 2022. 111005.
2. Zhang Z, Liu X, Zhao D, Post S, Chen J. Overview of the development and application of wind energy in New Zealand. Energy Built Environ.
2023;4(6):725–42. https:// doi. org/ 10. 1016/j. enbenv. 2022. 06. 009.
3. Hassan Q, etal. The renewable energy role in the global energy transformations. Renew Energy Focus. 2024;48: 100545. https:// doi. org/
10. 1016/j. ref. 2024. 100545.
4. Sabev E, Trifonov R, Pavlova G, Rainova K. Cybersecurity analysis of wind farm SCADA systems. In 2021 International Conference on
Information Technologies (InfoTech), IEEE, 2021, pp. 1–5. https:// doi. org/ 10. 1109/ InfoT ech52 438. 2021. 95485 89.
5. Farrar NO, Ali MH, Dasgupta D. Articial Intelligence and machine learning in grid connected wind turbine control systems: a compre-
hensive review. Energies. 2023;16(3):1530. https:// doi. org/ 10. 3390/ en160 31530.
6. Yildiz SS. Determining wind energy potential using geographic information system functions: a case study in Balıkesir, Türkiye. Appl Sci.
2023;13(16):9183. https:// doi. org/ 10. 3390/ app13 169183.
7. Yilmaz I. A hybrid DEA–fuzzy COPRAS approach to the evaluation of renewable energy: a case of wind farms in Türkiye. Sustainability.
2023;15(14):11267. https:// doi. org/ 10. 3390/ su151 411267.
8. Rediske G, Burin HP, Rigo PD, Rosa CB, Michels L, Siluk JCM. Wind power plant site selection: a systematic review. Renew Sustain Energy
Rev. 2021;148: 111293. https:// doi. org/ 10. 1016/j. rser. 2021. 111293.
9. Mansour S. Geospatial modelling of drought patterns in Oman: GIS-based and machine learning approach. Model Earth Syst Environ.
2024;10(3):3411–31. https:// doi. org/ 10. 1007/ s40808- 024- 01958-9.
10. Chowdhury MS. Comparison of accuracy and reliability of random forest, support vector machine, articial neural network and maximum
likelihood method in land use/cover classication of urban setting. Environ Challenges. 2024;14: 100800. https:// doi. org/ 10. 1016/j. envc.
2023. 100800.
11. Helmi AM, Elgamal M, Farouk MI, Abdelhamed MS, Essawy BT. Evaluation of geospatial interpolation techniques for enhancing spati-
otemporal rainfall distribution and lling data gaps in Asir Region, Saudi Arabia. Sustainability. 2023;15(18):14028. https:// doi. org/ 10.
3390/ su151 814028.
12. Rajamani T, Sevugan P, Ragupathi S. Automatic building footprint extraction and road detection from hyperspectral imagery. J Electron
Imaging. 2022. https:// doi. org/ 10. 1117/1. JEI. 32.1. 011005.
13. Jain R, Bhat A. Determining statistically signicant road accident spatial hotspots using machine learning approaches. In 2022 4th Inter-
national conference on advances in computing, communication control and networking (ICAC3N), IEEE, 2022, pp. 214–221. https:// doi.
org/ 10. 1109/ ICAC3 N56670. 2022. 10074 260.
14. Rezvani SM, Falcão MJ, Komljenovic D, de Almeida NM. A systematic literature review on urban resilience enabled with asset and disaster
risk management approaches and GIS-based decision support tools. Appl Sci. 2023;13(4):2223. https:// doi. org/ 10. 3390/ app13 042223.
15. Bin Abu Soan ADA, Lim HR, Siti Halimatul Munawaroh H, Ma Z, Chew KW, Show PL. Machine learning and the renewable energy
revolution: exploring solar and wind energy solutions for a sustainable future including innovations in energy storage. Sustain Dev.
2024;32(4):3953–78. https:// doi. org/ 10. 1002/ sd. 2885.
16. Fan Z, Yan Z, Wen S. Deep learning and articial intelligence in sustainability: a review of SDGs, renewable energy, and environmental
health. Sustainability. 2023;15(18):13493. https:// doi. org/ 10. 3390/ su151 813493.
17. Khan MN, Rashid MI, Ahmad I. Hybrid decision support system for wind farm site selection using fuzzy AHP and F-TOPSIS. Renew Energy.
2020;147:1255–71.
18. Cerna PD, etal. Wind power plant site selection using integrated machine learning and multiple-criteria decision making technique. E3S
Web Conf. 2023;405:02030. https:// doi. org/ 10. 1051/ e3sco nf/ 20234 05020 30.
19. Sotiropoulou KF, Vavatsikos AP, Botsaris PN. A hybrid AHP-PROMETHEE II onshore wind farms multicriteria suitability analysis using kNN
and SVM regression models in northeastern Greece. Renew Energy. 2024;221: 119795. https:// doi. org/ 10. 1016/j. renene. 2023. 119795.
Content courtesy of Springer Nature, terms of use apply. Rights reserved.

Vol:.(1234567890)
Research
Discover Computing (2025) 28:22 | https://doi.org/10.1007/s10791-025-09511-7
20. Gorsevski PV, Cathcart SC, Mirzaei G, Jamali MM, Ye X, Gomezdelcampo E. A group-based spatial decision support system for wind farm
site selection in Northwest Ohio. Energy Policy. 2013;55:374–85. https:// doi. org/ 10. 1016/j. enpol. 2012. 12. 013.
21. Xenitidis K, Ioannou K, Tsantopoulos G, Myronidis D. On the usage of articial neural networks for the determination of optimal wind
farms allocation. Sustainability. 2023;15(24):16938. https:// doi. org/ 10. 3390/ su152 416938.
22. Albraheem L, AlAwlaqi L. Geospatial analysis of wind energy plant in Saudi Arabia using a GIS-AHP technique. Energy Rep. 2023;9:5878–98.
https:// doi. org/ 10. 1016/j. egyr. 2023. 05. 032.
23. Demir G, Riaz M, Deveci M. Wind farm site selection using geographic information system and fuzzy decision making model. Expert Syst
Appl. 2024;255: 124772. https:// doi. org/ 10. 1016/j. eswa. 2024. 124772.
24. Demir A, Dinçer AE, Çiftçi C, Gülçimen S, Uzal N, Yılmaz K. Wind farm site selection using GIS-based multicriteria analysis with Life cycle
assessment integration. Earth Sci Informatics. 2024;17(2):1591–608. https:// doi. org/ 10. 1007/ s12145- 024- 01227-4.
25. Zhao H, Wang S, Lu C. A study on site selection of wind power plant based on prospect theory and VIKOR: a case study in China. Kyber-
netes. 2024. https:// doi. org/ 10. 1108/K- 01- 2024- 0022.
26. Gündüz S, Gündüz S. Determination of last glacial equilibrium line altitudes of Türkiye by glacial mass balance model. Master’s Thesis,
Avrasya Yerbilimleri Enstitüsü, 2018.
27. Royal Danish General Consulate of Denmark in Istanbul. Wind Energy Market in Türkiye. https:// tyrki et. um. dk/ en/-/ media/ count ry- sites/
tyrki et- en/ the- trade- counc il/ tc- Türkiye- windt easer- 2021. ashx.
28. Türkiye Map 360, 2024. https:// turke ymap3 60. com.
29. Kohyama S, Kohsaka R. Wind farms in contested landscapes: procedural and scale gaps of wind power facility constructions in Japan.
Energy Environ. 2024;35(3):1396–415. https:// doi. org/ 10. 1177/ 09583 05X22 11413 96.
30. Karamountzou S, Vagiona DG. Suitability and sustainability assessment of existing onshore wind farms in Greece. Sustainability.
2023;15(3):2095. https:// doi. org/ 10. 3390/ su150 32095.
31. Villacreses G, Gaona G, Martínez-Gómez J, Jijón DJ. Wind farms suitability location using geographical information system (GIS), based
on multi-criteria decision making (MCDM) methods: the case of continental Ecuador. Renew Energy. 2017;109:275–86. https:// doi. org/
10. 1016/j. renene. 2017. 03. 041.
32. El-hosiny AM, El-Ghareeb H, Shabana BT, AbouElfetouh. Sustainable energy production’s spatial determination framework, based on
multi criteria decision making and geographic information system under neutrosophic environment: a case study in Egypt. Neural Netw
Stat. 2021;42:136–156.
33. Değirmenci SC, Bingöl S, Sofuoglu F. MCDM analysis of wind energy in Türkiye : decision making based on environmental impact. Environ
Sci Pollut Res. 2018;25:19753–66.
34. Ekeanyanwu CV, Obisakin IF, Aduwenye P, Dede -Bamfo N. Merging GIS and machine learning techniques: a paper review. J Geosci Environ
Prot. 2022;10(09):61–83. https:// doi. org/ 10. 4236/ gep. 2022. 109004.
35. Mitchell TM. Does machine learning really work? AI Mag. 1997;18(3):11.
36. Hassan LH, Moghavvemi M, Almurib HAF, Steinmayer O. Current state of neural networks applications in power system monitoring and
control. Int J Electr Power Energy Syst. 2013;51:134–44. https:// doi. org/ 10. 1016/j. ijepes. 2013. 03. 007.
37. Rahul K, Banyal RK, Goswami P, Kumar V. Machine learning algorithms for big data analytics. 2021, pp. 359–367. https:// doi. org/ 10. 1007/
978- 981- 15- 6876-3_ 27.
38. Tang Y, Chang Y, Li K. Applications of K-nearest neighbor algorithm in intelligent diagnosis of wind turbine blades damage. Renew Energy.
2023;212:855–64. https:// doi. org/ 10. 1016/j. renene. 2023. 05. 087.
39. Shahiri AM, Husain W, Rashid NA. A review on predicting student’s performance using data mining techniques. Procedia Comput Sci.
2015;72:414–22. https:// doi. org/ 10. 1016/j. procs. 2015. 12. 157.
40. Lee S, Chung JY. The machine learning-based dropout early warning system for improving the performance of dropout prediction. Appl
Sci. 2019;9(15):3093. https:// doi. org/ 10. 3390/ app91 53093.
41. Fahd K, Miah SJ, Ahmed K. Predicting student performance in a blended learning environment using learning management system
interaction data. Appl Comput Informatics. 2021. https:// doi. org/ 10. 1108/ ACI- 06- 2021- 0150.
42. Rodriguez-Galiano VF, Ghimire B, Rogan J, Chica-Olmo M, Rigol-Sanchez JP. An assessment of the eectiveness of a random forest clas-
sier for land-cover classication. ISPRS J Photogramm Remote Sens. 2012;67:93–104. https:// doi. org/ 10. 1016/j. isprs jprs. 2011. 11. 002.
43. Du Ke-Lin M, Swamy NS. Support vector machines. In: Du Ke-Lin M, Swamy NS, editors. Neural networks and statistical learning. London:
Springer London; 2014. p. 469–524. https:// doi. org/ 10. 1007/ 978-1- 4471- 5571-3_ 16.
44. Ovirianti NH, Muhammad Zarlis HM. Support vector machine using a classication algorithm. J Dan Penelit Tek Inform. 2022;7:2103–7.
45. Viet TN, Le Minh H, Hieu LC, Anh TH. The Naïve Bayes algorithm for learning data analytics. Indian J Comput Sci Eng. 2021;12(4):1038–43.
https:// doi. org/ 10. 21817/ indjc se/ 2021/ v12i4/ 21120 4191.
46. Yang F-J. An implementation of Naive Bayes Classier. International conference on computational science and computational intelligence
(CSCI), IEEE, 2018, pp. 301–306. https:// doi. org/ 10. 1109/ CSCI4 6756. 2018. 00065.
Publisher’s Note Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional aliations.
Content courtesy of Springer Nature, terms of use apply. Rights reserved.

1.
2.
3.
4.
5.
6.
Terms and Conditions
Springer Nature journal content, brought to you courtesy of Springer Nature Customer Service Center GmbH (“Springer Nature”).
Springer Nature supports a reasonable amount of sharing of research papers by authors, subscribers and authorised users (“Users”), for small-
scale personal, non-commercial use provided that all copyright, trade and service marks and other proprietary notices are maintained. By
accessing, sharing, receiving or otherwise using the Springer Nature journal content you agree to these terms of use (“Terms”). For these
purposes, Springer Nature considers academic use (by researchers and students) to be non-commercial.
These Terms are supplementary and will apply in addition to any applicable website terms and conditions, a relevant site licence or a personal
subscription. These Terms will prevail over any conflict or ambiguity with regards to the relevant terms, a site licence or a personal subscription
(to the extent of the conflict or ambiguity only). For Creative Commons-licensed articles, the terms of the Creative Commons license used will
apply.
We collect and use personal data to provide access to the Springer Nature journal content. We may also use these personal data internally within
ResearchGate and Springer Nature and as agreed share it, in an anonymised way, for purposes of tracking, analysis and reporting. We will not
otherwise disclose your personal data outside the ResearchGate or the Springer Nature group of companies unless we have your permission as
detailed in the Privacy Policy.
While Users may use the Springer Nature journal content for small scale, personal non-commercial use, it is important to note that Users may
not:
use such content for the purpose of providing other users with access on a regular or large scale basis or as a means to circumvent access
control;
use such content where to do so would be considered a criminal or statutory offence in any jurisdiction, or gives rise to civil liability, or is
otherwise unlawful;
falsely or misleadingly imply or suggest endorsement, approval , sponsorship, or association unless explicitly agreed to by Springer Nature in
writing;
use bots or other automated methods to access the content or redirect messages
override any security feature or exclusionary protocol; or
share the content in order to create substitute for Springer Nature products or services or a systematic database of Springer Nature journal
content.
In line with the restriction against commercial use, Springer Nature does not permit the creation of a product or service that creates revenue,
royalties, rent or income from our content or its inclusion as part of a paid for service or for other commercial gain. Springer Nature journal
content cannot be used for inter-library loans and librarians may not upload Springer Nature journal content on a large scale into their, or any
other, institutional repository.
These terms of use are reviewed regularly and may be amended at any time. Springer Nature is not obligated to publish any information or
content on this website and may remove it or features or functionality at our sole discretion, at any time with or without notice. Springer Nature
may revoke this licence to you at any time and remove access to any copies of the Springer Nature journal content which have been saved.
To the fullest extent permitted by law, Springer Nature makes no warranties, representations or guarantees to Users, either express or implied
with respect to the Springer nature journal content and all parties disclaim and waive any implied warranties or warranties imposed by law,
including merchantability or fitness for any particular purpose.
Please note that these rights do not automatically extend to content, data or other material published by Springer Nature that may be licensed
from third parties.
If you would like to use or distribute our Springer Nature journal content to a wider audience or on a regular basis or in any other manner not
expressly permitted by these Terms, please contact Springer Nature at
onlineservice@springernature.com