19(2): 81-87, 2024 www.thebioscan.

com

ADVANCING SOIL NUTRIENT MANAGEMENT IN AGRICULTURE WITH

INTEGRATING MACHINE LEARNING AND FUZZY LOGIC
APPROACHES
GIGI ANNEE MATHEW1, VARSHA JOTWANI2 AND A. K. SINGH*3
1
Research Scholar, Department of Computer Science & IT, Rabindranath Tagore University, Bhopal-464993, India,
E-mail: gigiannee@gmail.com
2
HoD (CS & IT), Department of Computer Science & IT, Rabindranath Tagore University,Bhopal-464993, India,
E-mail: varsha.jotwani@aisectuniversity.ac.in
3
Scientist, Krishi Vigyan Kendra, Jawaharlal Nehru Krishi Vishwa Vidyalaya, Jabalpur- 482004, M.P., India,
E-mail: singhak123@rediffmail.com,
ORCID ID: 0000-0002-7644-5802
Corresponding Author Contact Number: +919424638238

DOI: https://doi.org/10.63001/tbs.2024.v19.i02.pp81-87

KEYWORDS ABSTRACT
Soil nutrient
management, Agriculture faces multifaceted challenges that affect food security, sustainability, and economic growth. Soils
machine learning, serve as the foundation for food production. However, indiscriminate use of fertilizers has led to soil pollution
fuzzy logic, and degradation, necessitating integrated nutrient management practices. Machine learning (ML) emerges as a
classification, transformative technology in agriculture, offering solutions across various domains. Fuzzy logic, with its ability
categorization, to handle uncertainty and imprecision, complements machine learning in agricultural decision support systems.
prediction, This paper explores the utilization of fuzzy logic for soil nutrient categorization and decision-making, along with
recommendation system
the analysis of machine learning models for predicting soil fertility and crop yield and also examines the
Received on:
recommendation of suitable crops and fertilizers based on soil characteristics. These models leverage diverse
05-04-2024 algorithms such as K- Nearest Neighbours, Random Forest, Naive Bayes, Support Vector Machine, Decision
Trees and ensemble classifiers to offer accurate predictions and recommendations. The integration of ML and
Accepted on: fuzzy logic in agriculture represents a potential approach to tackling agricultural challenges, advancing
sustainable soil management practices, and elevating crop productivity.
02-09-2024

Corresponding author

Problems in Agriculture unmanaged (Savary et al., 2019). Moreover, the loss of

Global and local agriculture confronts a number of biodiversity due to agricultural intensification,
formidable obstacles that have an effect on food security, monoculture practices, and habitat destruction
sustainability, and economic growth. Severe conditions undermines ecosystem services critical for agricultural
including heat waves, floods, and droughts brought on by resilience and long-term sustainability (Tscharntke et al.,
climate change reduce animal production and yields of 2012).
crops. (Lobell et al., 2011). Water scarcity intensifies Rural poverty and food insecurity are a persistent
agricultural problems, particularly in arid and semi-arid challenge, particularly among small farmers in developing
regions. As demand for water increases, its resources nations, who often lack access to markets, credit,
decline due to pollution, overuse, and the effects of technology, and resources necessary for sustainable
climate change (FAO, 2020). Soil degradation presents agricultural practices (FAO, 2021). Food wastage and
another concern, with erosion, salinization, acidification, losses in the agricultural supply chain further aggravate
and loss of fertility posing threats to agricultural global hunger, economic losses, and environmental
productivity and sustainability (Lal, 2015). Pests and degradation, highlighting inefficiencies in production,
diseases continually affect crops and livestock, causing distribution, and consumption (FAO, 2019).
substantial losses in yield and economic returns, if left

81
Soil Nutrient Management et al. (2021) propose a soil fertilization system that
Soil is one of the most important resource of agriculture. It merges evolutionary algorithms with soil nutrient data and
is estimated that 95% of food is directly or indirectly crop requirements. Mahagaonkar (2019) focuses on crop
produced on soils. Soils supply the essential nutrients, yield prediction and fertilizer recommendation using
water, oxygen and root support that our food-producing machine learning algorithms. Thus, a recommendation
plants need to grow and flourish (FAO, 2015), The quality system for farmers on integrated soil nutrient
and quantity of food produced solely depend on the soil management can be developed for specific crops based on
health (Lal, 2015). An increase in food demand has led to the assessed soil nutrient status, thereby improving soil
the application of large quantities of fertilizers, resulting health and crop production (Mella and Venkata, 2022).
in pollution and degradation of the soil (Juhi Reshma et Machine Learning
al., 2021). In the majority of areas, farmers are applying Machine learning (ML) is a process by which computers
chemical fertilizers without soil testing which leads to a acquire the ability to execute tasks similar to those carried
depletion of soil health. Deficient nutrients impact the out by humans. Typically, ML empowers systems to
growth of plants, plant physiological disorders and yield autonomously improve and acquire knowledge from
(Smith et al., 2022). Integrated nutrient management experience, without explicit programming. It is one of the
(INM) is the sustainable solution for healthy soil. INM plays leading technology developments in the current era, often
a vital role in maintaining soil productivity, ensuring associated with the fourth industrial revolution (Sarker,
productive and sustainable agriculture, reducing inputs 2021). ML is a subset of Artificial Intelligence (AI)
cost by using farm wastes, green manures, biofertilizers dedicated to learning. It enhances performance prediction
and growing leguminous crops (Balusamy et al., 2020). by analyzing multiple features, identifying patterns,
Soil nutrient deficiency in Indian soil and its associated correlations, and insights from datasets and past
impact on crop yield has been widely acknowledged and experiences. Numerous models are constructed with
reported (Pandey et al., (2016). Numerous policy various features and parameters based on historical data.
initiatives have been taken both at the State and National Testing evaluates performance using unused historical data
level. The soil health card introduced by the Government (Klompenburg et al. 2020).
of India in 2015, is the best example. Though awareness “Machine learning" was coined by Arthur Samuel in 1959,
about this initiative among farmers is high, only a less pioneering AI in computer gaming. It evolved from early
followed the recommended rate of fertilizer application. chess programmes in the 1940s to the development of
Some of the emphasized problems are the delayed receipt neural networks and deep learning. This was due to the
of soil test reports after soil sample submission, shortage significant contributions of pioneers like Nilsson, Duda,
of soil experts and accessibility (Reddy, A. A. 2019). Hart, Hinton, Ng, Bengio, and LeCun, who laid the
The accuracy of soil testing outcomes and foundation of modern neural network technologies
recommendations for fertilizers is reliant upon the quality (Sharma et al., 2021). Three key phases of ML are data
of soil sampling (Ministry of Agriculture and Farmers’ input, model construction, and generalization (Fig 1).
Welfare, 2020). A fertilizer recommendation system based ‘Generalization’ entails predicting outcomes for inputs not
on soil analysis should be designed to ensure widespread included in the algorithm's training data. However, ML
adoption by farmers (Reddy, A. A. 2019). A comprehensive algorithms demonstrate proficiency in addressing complex
approach to agricultural management can be achieved by challenges where human expertise may be insufficient, for
integrating various computational techniques. Arogundade example, weather prediction, spam identification, plant
et al. (2021) present a system utilizing fuzzy logic for disease diagnosis, pattern detection, etc. (Sharma et al.,
soil fertility analysis and decision-making, while Ahmed 2020).

Dataset
Model building
(Pre-processed Generalization
using ML algm
data)

Fig 1: A Machine learning process

Classification of Machine learning techniques unfold the underlying structure or patterns
Machine learning algorithms can be mainly classified as in the data, for example, clustering similar
supervised, unsupervised and reinforcement learning data points together or reducing the
paradigms. These serve tasks including classification, dimensionality of the data.
regression, clustering, and decision making, with diverse iii. Reinforcement learning is a type of learning
algorithms designed to suit each task's requirements and where an agent learns to interact with an
objectives. ML classification is important as it provides a environment to achieve a goal or maximize
framework that enables the selection of appropriate a cumulative reward. The agent takes
algorithms tailored to the specific nature of the data and actions in the environment, receives
the desired outcome (Sharma et al., 2020 and Liakos et feedback in the form of rewards or
al., 2018) as shown in Table 1. penalties, and adjusts its behaviour over
time to maximize its cumulative reward
i. Supervised learning, the algorithm is (Sharma et al., 2021).
trained on a labelled dataset, where each
example in the dataset is associated with a
label or outcome. Here the goal is for the
algorithm to learn a mapping from input
data to output labels, enabling it to make
predictions or decisions on new, unseen
data.
ii. Unsupervised learning, the algorithm is
given input data without any explicit labels
or outcomes. The algorithm's task is to

82
 Table 1: Categorization of ML algorithms
Machine learning is revolutionizing agriculture
Category Task Algorithms with its diverse applications. Crop yield prediction
involves analyzing historical data, including
- Decision Trees (DT)
weather patterns and soil conditions, to inform
- Random Forests (RF) farmers about resource allocation and harvest
planning. Disease and pest detection, facilitated
Classification - k-Nearest Neighbours (KNN) by machine learning models trained on images of
- Support Vector Machines (SVM) healthy and diseased crops, enables real-time
identification of issues, reducing crop losses
Super - Naive Bayes (NB) through timely intervention. Weed identification
vised and management benefit from machine learning
learni -Neural Networks (e.g., Multi-layer Perceptron)
algorithms' ability to differentiate between crops
ng - Gradient Boosting Machines (e.g., XGBoost) and weeds, allowing for targeted herbicide
- Linear Regression application and minimizing environmental impact.
Precision agriculture techniques, driven by
- Polynomial Regression machine learning technologies like Variable Rate
Regression Technology (VRT), optimize resource use such as
- Ridge Regression
water, fertilizers, and pesticides, resulting in
- Lasso Regression improved crop yields. Additionally, machine
learning recommendation systems process diverse
- Support Vector Regression
data sources to offer personalized insights, from
- Decision Trees regression optimal planting times to irrigation schedules,
empowering farmers with informed
- Random Forests regression
decision-making. Machine learning also extends
its impact on livestock management, addressing
Unsupervised Clustering - K-Means Clustering animal welfare and enhancing livestock
production practices (Liakos et al., 2018). The
learning - Agglomerative Hierarchical Clustering success of a machine learning solution depends on
the type of data and the performance of learning
- Density-Based Spatial Clustering of algorithms, which include classification,
Applications with Noise (DBSCAN) regression, clustering, feature engineering,
- Gaussian Mixture Models (GMM) dimensionality reduction, association rule
learning, and reinforcement learning techniques
- Mean Shift Clustering (Sarker, 2021).
Fuzzy Logic
Fuzzy logic is a mathematical system that
- Fuzzy Clustering addresses uncertainty and imprecision like human
cognition. It was developed by LotfiZadeh in the
Reinforcement - Q-Learning 1960s as an extension of classical (crisp) logic,
learning which deals with binary values (true or false).
Decision - R Learning Conversely, fuzzy logic permits varying degrees of
making truth; something can be partially true or partially
- Q Learning false.
In traditional set theory, an element either
- TD Learning belongs to a set or not. Fuzzy sets, however,
introduce the concept of partial membership. This
means that instead of being strictly in or out of a
Application of Machine Learning in Agriculture set, elements can belong to a set to a certain degree,
Machine learning finds applications in diverse areas, typically represented by a value between 0 and 1.
including stock price prediction, scientific research, Membership functions define how elements are assigned
marketing strategies, cyber security, e-commerce, mobile degrees of membership in fuzzy sets. These functions map
data processing, health analytics, user modelling, the values of variables to their corresponding membership
behavioural analytics, and clustering. These applications degrees. Membership functions can take various shapes,
leverage historical data to make predictions and derive such as triangular, trapezoidal, or Gaussian, depending on
insights across various domains (Sharma et al. 2021) and the nature of the linguistic variable and the problem
(Sarker2021). Modern agriculture is also in the process of domain. Linguistic variables are variables whose values are
employing machine learning techniques. Machine learning expressed in natural language terms rather than numerical
can benefit farmers to minimize the losses in farming by values. Each linguistic variable is associated with a set of
providing recommendations and insights about the crops linguistic terms (e.g., "Low", "Medium", "High" for
he is growing. Machine learning can be applied in "Temperature") and corresponding membership functions
agriculture to alleviate the problems in the three areas of that specify the degree of membership for each term
pre-harvesting, harvesting and post-harvesting. The (Onashoga et al. 2018).
application of machine learning in agriculture allows more Fuzzification is the task of partitioning the input variable
efficient and precise farming with less manpower for high- and assigning the linguistic label for each partition. It
quality production (Meshram, 2021). transforms discrete into a continuous form by linguistic
variables and membership functions of a fuzzy set rules and input data to produce a fuzzy output, which
(Prabakaran et al., 2018). Fuzzy logic systems operate on represents the truth values of possible output values.
"if-then" rules defined using linguistic variables, capturing When a precise output is needed, defuzzification is
expert knowledge and heuristics to make decisions based applied as shown in Fig. 2 (Arogundade et al., 2021).
on input data. The fuzzy inference process combines these

83
 optimizing crop production. Choudhary et al. (2022)
introduce a fuzzy inference system designed to predict
crop yield based on soil fertility parameters using MatLab.
The soil dataset is obtained from Kaggle having attributes
such as levels of N, P, K and pH, temperature, humidity,
and rainfall. The study aims to address the challenges of
uncertainty and variability in agricultural production by
leveraging fuzzy logic principles. Through the integration
of soil fertility data and fuzzy inference techniques, the
authors develop a model capable of providing accurate
predictions of crop yield.
Pant et al. (2020) developed a fuzzy control system for
assessing the quality and fertility of 106 soil samples in the
Nainital district of Uttarakhand, India. The study focused on
Fig 2: Fuzzy logic system developing rule descriptions based on N, P, and K to evaluate
Fuzzy logic finds applications in fields such as control soil parameters and determine their impact on soil health and
systems, artificial intelligence, and machine learning, fertility levels. Fuzzy logic principles were employed using
particularly when handling uncertain or ambiguous data, Python’s skfuzzy tool to create a framework that integrates
enabling a more human-like approach to decision-making. various soil quality indicators and expert knowledge. Ogunleye
It is utilized for making decisions when faced with et al (2018) proposed a fuzzy logic-based tool to offer
uncertainty Alfin et al. (2017) and Jane and Ganesh accurate and accessible predictions of soil fertility levels in
(2019). The fuzzy logic model forms the core of the their study in Nigeria. Through data collection, preprocessing,
decision support system, which provides recommendations and the formulation of fuzzy logic models incorporating
on fertilizer applications based on real-time or historical linguistic variables and rules, the system evaluated input soil
data input by the user. The system considers the parameters to determine soil fertility. The inference engine
uncertainty and imprecision inherent in agricultural data computed the degree of membership for each linguistic
and provides flexible and adaptive recommendations. By variable, resulting in an output indicating the inferred soil
using the fuzzy decision support system, farmers can fertility level. The system aimed to assist farmers, land
improve crop productivity by applying the right amount of managers, and policymakers in making informed
fertilizer at the right time and in the right place decisions regarding soil nutrientmanagement practices by
(Prabakaran et al., 2018). providing accurate predictions of soil fertility, ultimately
Fuzzy Logic for Soil Nutrient Categorization and Decision- contributing to enhanced agricultural productivity and
Making sustainability in Nigeria and similar regions. Wan Yahya et al.
Soil nutrient parameters are pH (reaction of soil), EC (2021) explored the application of fuzzy logic in predicting
(electrical conductivity), OC (organic carbon), N the micronutrient requirements of Harumanis mango across
(nitrogen), P (phosphorus), K (potassium), S (sulphur), Zn various growth stages. They developed a predictive model
(zinc), B (Boron), Fe (iron), Mn (manganese), and Cu based on 27 fuzzy rules of N, P, K present as input and N, P, K
(copper). Soil nutrients have been classified as Physical needed as output using MatLab. Through experimentation and
(pH, EC, OC), Macronutrients (N, P, K), Secondary Nutrients validation, they demonstrate the efficacy of their fuzzy logic-
(S), and Micronutrients (Fe, Zn, Cu, Mn, B). The nutrient based approach in providing reliable predictions of
status of the soil thus can be determined based on the micronutrient demand, thereby offering a valuable tool for
physical, macronutrients, secondary nutrients and optimizing nutrient management strategies in mango
micronutrients of that soil (Paul et al., 2015). cultivation to enhance yield and quality.
Soil nutrient categorization classifies soils into qualitative Machine learning for Prediction and Recommendation
categories based on the scale of optimum nutrient levels System Based on Soil Nutrients
or indices. The category explains the nutrient status of the Classification is supervised machine learning that instructs
particular soil to the readings of nutrients of that soil as machines on how to group data based on specific criteria,
Low, Medium, or High. Traditional soil nutrient assessment such as predetermined characteristics. It predicts a data
methods often provide precise numerical values for point that belongs to one of the predefined classes. The
nutrient concentrations. These values may not fully capture prediction is based on learning from a known dataset (Kotu
the inherent imprecision associated with soil properties and Deshpande, 2019). It is a two-step process, by
and measurements. Fuzzy logic allows for the use of analyzing the attributes within the dataset to build a
linguistic variables and fuzzy sets torepresent imprecise model with predetermined class labels and then estimate
information more intuitively. The soil nutrient values can the analytical accuracy of the built-in model. Before
be converted to soil nutrient status using a fuzzy if-then applying classification techniques, it is crucial to perform
rule method on a set range (Bang et al., 2019). Soil data cleaning, data transformation, and relevance analysis
nutrient status is the backbone on which all input-based to refine the dataset and optimize its suitability for
high agricultural production systems can be built (Lal, accurate model training and prediction. The results of the
2015). classification models are evaluated according to its
Onashoga et al. (2018) presented a fuzzy logic-based scalability, speed, predictive accuracy, robustness and
decision support system design to aid in soil selection for interpretability (Awasthiand Bansal, 2017).
vegetable cultivation. The system addresses the challenges Pandith et al. (2020) explained the prediction of the
of soil variability and complexity in agricultural decision- Mustard crop yield model based on soil analysis of a
making processes. By utilizing fuzzy logic principles, the dataset consisting of 5000 instances with 11 input
authors develop a framework that integrates various soil parameters representing soil nutrient levels of the Jammu
parameters and expert knowledge to recommend suitable region by employing machine learning algorithms like K-
soil types for specific vegetable crops. The system's Nearest Neighbor (KNN), Naive Bayes, Multinomial Logistic
usefulness is proven against conventional methods, Regression (MLR), Artificial Neural Network (ANN) and
showcasing its ability to provide accurate and intuitive Random Forest (RF). Each model's performance assessed
recommendations for soil selection in vegetable farming. was based on metrics such as accuracy, recall, precision,
Fuzzy logic offers agricultural decision support systems for specificity and F-Score by dividing the dataset into 70:30
the suitable crop to be sowed based on soil management training and testing sets respectively.
for improved yield. Fuzzy logic can be applied in
predicting crop yield based on soil fertility parameters,
thereby enhancing agricultural decision-making and

84
Based on comparative analysis, KNN and RF predicted the Among these, the Random Forest algorithm demonstrated
highest accuracy of 88.67 and 94.13 percent respectively. the highest accuracy of 95% for the given dataset. The
MLR – 80.24%, ANN – 76.86% and Naive Bayes predicted the proposed system incorporates a mobile phone as the end-
lowest accuracy of 72.33%. Bondre and Mahagaonkar user application, utilizing GPS for location identification.
(2019) designed a system to predict crop yield using Users input their land area and soil type, enabling the
machine learning algorithms such as Support Vector system to recommend suitable crops for cultivation. Gosai
Machine (SVM) and Random Forest (RF) based on five years et al. (2021) proposed a crop suggestion system that
of prior agricultural data. The system also suggests proposes the most suitable crop by predicting the accuracy
appropriate fertilizers for eachspecific crop. The of the future production of different crops based on
performance evaluation was done on precision, f1-score various parameters (N, P, K, soil pH, humidity,
recall, and an average value. Yield prediction was temperature, and rainfall). The data sensed by the sensors
conducted using crop yield data, nutrient information, was stored on the microcontroller and analyzed by Naive
and location data. Additionally, fertilizer Bayes, Support Vector Machine, Decision Trees, Logistic
recommendations were made based on fertilizer data, Regression, Random Forest, and XGBoost. The best
specific crop details, and location data. RF had good accuracy is achieved by XGBoost (99.31 %).
accuracy (86.35%) when compared with SVM for soil Various machine learning models used for prediction and
classification. The SVM algorithm exhibits 99.47% accuracy their effectiveness for agricultural use based on soil
for yield prediction, while the RF algorithm's accuracy was nutrients are described in Table 2.
97.48%.
Archana and Saranya (2020) designed a comprehensive Table 2: Overview of prediction models and its performance
system for crop rotation, yield prediction, and forecasting for agriculture use
along with fertilizer recommendations. The system Focus Methods Performance Reference
integrated N, P, K levels, soil type, soil texture, land type, Evaluation
pH, and soil EC as inputs. Moreover, the system utilized Prediction of Mustard Accuracy, KNN and RF had Pandith et al.
temperature data to offer the most suitable crop cropyield based on soil Recall, the highest (2020)
suggestions. The proposed system was developed using analysisusing KNN, Precision, accuracy (88.67%
voting based ensemble classifier algorithm. An ensemble Naive Bayes, MLR, Specificity, and
ANN, and RF F-Score 94.13%
classifier is a cluster of classifiers of Naive Bayes, Random
respectively),
Forest and Chi-square automatic interaction detection while Naive
(CHAID). An output of each classifier suggests crops Bayes had the
separately for the user given input suggests the crops with lowest(72.33%).
the highest vote are to use. The yield calculation for the Prediction of crop yield Precision, Random Forest Bondre et al.
crop suggested by the Ensemble Classifier model is andfertilizer Recall, F1- outperformedSVM (2019)
computed relying on the area of the cultivation land in recommendation using Score, in soil
acres. The distance measure was applied to the provided SVM and RF Averag classification
e (86.35% vs. 99.47%
NPK values to generate the top three alternate crop accuracy).
suggestions. Performance is determined based on response
time using a stopwatch and accuracy metric by the Design of a crop Respons The proposed Archana and
formula: Number of correct predictions/ Total number of recommendation eTime, system Saranya
predictions made. According to the authors, the proposed system using Accurac achieved 92% (2020)
crop recommendation system gives 92% accuracy. ensemble classifiers yMetric accuracy in
crop
Malik et al. (2021) performed an analytical comparison of recommenda
soil characteristics for the prediction of fertility and crop tions.
productivity employing machine learning techniques K- Comparative analysis Accuracy KNN achieved Malik et
Nearest Neighbour algorithm, Naive Bayes algorithm and of soil properties using 91.179%, Naive al.(2021)
Decision Trees classifier. The dataset of 1320 cases and KNN, Bayes 76.426%,
four parameters (temperature, pH, moisture and sunlight Naive Bayes, and and
of) potato, tomato, and chilli crops were taken for the Decision Decision Trees
95.361%
research and implementation was conducted using Python.
Trees accuracy.
The visualization of data was accomplished using the
Matplotlib library in Python. KNN algorithm predicted the
yield by considering the Euclidean distance, with an
accuracy of 91.179%. The naive Bayes algorithm predicted
using the Bayes theorem, with a relatively lower accuracy Prediction of soil MSE, Cross- Decision Tree Rajamani
of 76.426%. Decision Trees predicted the yield through the fertility using DT, validation, achieved 99% ckam
Gini index, with the utmost prediction of 95.361%. KNN, SVM Accuracy accuracy with and
the lowest Savitha
However, the soil fertility prediction work was carried out
MSE. (2021)
by Rajamanickam and Savitha (2021), by employing
Crop recommendation Accuracy RF achieved Pande et
Decision Tree (DT), K-Nearest Neighbour (KNN), and and yield prediction 95% accuracy al.(2021)
Support Vector Machine (SVM). The dataset (pH, EC, OC, using for crop
N, P, K, S, Cu, Fe, Zn, Mn) of 1000 soil samples was ANN, SVM, MLR, RF, recommendati
classified into two class labels: Ideal and NotIdeal. The KNN on.
dataset was split into training and testing sets with a ratio
of 70:30. Implementation was done using R Tool version CONFLICT OF INTEREST: None
3.5.3. The performance analysis was conducted using
various evaluation metrics, such as mean absolute error ETHICS: Not applicable
(MSE), cross- validation, and accuracy. DT resulted in 99%
accuracy with minimum MSE i.e., 0.01.
Pande et al., (2021) suggested a model proposing • Ahmed, U., Lin, J.C., Srivastava, G. and Djenouri,
profitable crop options for a specific location and Y. 2021. A nutrient recommendation system for soil
projecting yield based on cultivated area, GPS, and soil fertilization based on evolutionary computation.
type. They utilized various machine learning techniques Computers and Electronics in Agriculture.189: 1-7.
for yield projection, including ANN, SVM, Multivariate • Alfin, A.A. and Sarno, R. 2017. Soil irrigation fuzzy
Linear Regression, Random Forest, and KNN, tested on estimation approach based on decision making in
historical datasets from Maharashtra and Karnataka states. Sugarcane industry. In:11th International Conference

85
 on Information & Communication Technology and 996.
System (ICTS), Surabaya, Indonesia, 2017, pp. 137- • Liakos KG, Busato P, Moshou D, Pearson S, Bochtis
142. D.2018 Machine Learning in Agriculture: A Review.
• Amruta, N., Maruthi, J. B., Sarika, G. and Deepika, Sensors (Basel).18(8):2674
C. 2015. Effect of integrated nutrient management • Lobell, D. B., Schlenker, W., & Costa-Roberts, J.
and spacing on growth and yield parameters of (2011). Climate trends and global crop production
blackgram cv. LBG 625 (Rashmi). The Bioscan.10(1): since 1980. Science, 333(6042), 616-620.
193-198. • Malik, P., Sengupta, S. and Jadon, J.S. 2021.
• Archana, K. and Saranya, K.G. 2020. Crop yield Comparative analysis of soil properties to predict
prediction, forecasting and fertilizer recommendation fertility and crop yield using machine learning
using voting-based ensemble classifier. SSRG Int. J. algorithms. In: 2021 11th International Conference on
Comput. Sci. Eng. 7(2020): 1-4. Cloud Computing, Data Science & Engineering
• Arogundade, Oluwasefunmi & Atasie, C. & Misra, (Confluence). IEEE, 2021.
Sanjay & Sakpere, Aderonke & Abayomi-Alli, Olusola
• Mella, N.V.V.P. and Venkata, M.P. 2022. Crop yield
& Adesemowo, A. (2021). Improved Predictive System
prediction and Fertilizer Recommendation using
for Soil Test Fertility Performance Using Fuzzy Rule
Voting Based Ensemble Classifier.Journal of
Approach. 10.1007/978-981-16-0708-0_21. Awasthi,
Engineering Sciences. 13(8): 262-270.
N. and Bansal, A. 2017. Application of Data Mining
• Meshram, V.A., Patil, K., Meshram, V.V., Hanchate,
Classification Techniques on Soil Data using R.
D.B., &Ramkteke, S.D.(2021). Machine learning in
International Journal of Advances in Electronics and
agriculture domain: A state-of-art survey. Artificial
Computer Science. 4(2017) 33-37.
Intelligence in the Life Sciences1.
• Balusamy, A., Ramesh T., Choudhury, B.U.,
• Ogunleye GO, Fashoto SG, Mashwama P, Arekete SA,
Hazarika, S., Moirangthem, P., Chanu, L.J.,
Olaniyan OM, Omodunbi BA. 2018 Fuzzy Logic Tool to
Chakraborty, M.R.K., Laha R. and Rajappa, J.J.
Forecast Soil Fertility in Nigeria. Scientific World
2020. Acid soil health management for livelihood
Journal, Volume 2018
improvement. Extension bulletin no. 83. ICAR
• Onashoga, A., Ojesanmi, O., Johnson, F., Femi,
Research Complex for NEH Region, Umiam, Meghalaya -
E.A.2018. A fuzzy-based decision support system for
793103.
soil selection in olericulture. Journal of Agricultural
• Bang, S., Bishnoi, R., Chauhan, A.S., Dixit, A.K. and Informatics.9(3): 65-77.
Chawla, I. 2019. Fuzzy Logic based Crop Yield • Pande, S.M., Ramesh, P.K., Anmol, A., Aishwarya,
Prediction using Temperature and Rainfall parameters B.R., Rohilla, K. and Shaurya,
predicted through ARMA, SARIMA, and ARMAX models. • K. 2021. Crop Recommender System Using Machine
In: Twelfth International Conference on Learning Approach. In: 15th International Conference
Contemporary Computing (IC3), Noida, India, 2019. on Computing Methodologies and Communication
• Bondre, D.A. and Mahagaonkar, S. 2019. Prediction (ICCMC), Erode, India, 2021, pp.1066-1071
of crop yield and fertilizer recommendation using • Pandey, A., Patra, A. K., & Singh, D. (2016). Soil
machine learning algorithms. International Journal of nutrient status and crop yields in India: Importance of
Engineering Applied Sciences and Technology. 4(5): micro-nutrients. International Journal of Plant & Soil
371-376. Science, 11(2),1-14.
• Choudhary, M., Dutta, S. and Mahajan, J.K. 2022. • Pandith, V., Kour, H., Singh, S., Manhas, J. and
Fuzzy Inference system for soil fertility crop Sharma, V. 2020. Performance evaluation of machine
prediction. International Journal of Innovative learning techniques for mustard crop yield prediction
Research in Technology. 9(7): 257-260. from soil analysis. Journal of Scientific Research.
• Food and Agriculture Organization 2015. 2015 64(2): 394-398.
International Year of Soils. https://www.fao.org/soils- • Pant, H., Chandra, L., Bhatt, A., Pant, J. and Singh,
2015 M.K. 2020. Rule Descriptions for Soil Quality and Soil
• Food and Agriculture Organization 2019. The State Fertility Assessment using Fuzzy Control System.
of Food Security and Nutrition in the World 2019. International Journal of Recent Technology and
Retrieved from Engineering.8(6): 1341-1346.
http://www.fao.org/3/ca6030en/CA6030EN.pdf • Paul, M., Vishwakarma, S.K. and Ashok Verma, A.
• Food and Agriculture Organization 2020. The State 2015. Analysis of soil behavior and prediction of crop
of Food and Agriculture 2020. Retrieved from yield using data mining approach. In: International
http://www.fao.org/3/ca9647en/CA9647EN.pdf Conference on Computational Intelligence and
• Food and Agriculture Organization 2021. The State Communication Networks (CICN). 766-771.
of Food Security and Nutrition in the World 2021. • Prabakaran, G., Vaithiyanathan, D. and Ganesan, M.
Retrieved from 2018. Fuzzy decision support system for improving the
http://www.fao.org/3/ca9692en/CA9692EN.pdf crop productivity and efficient use of fertilizers.
• Gosai, D., Raval, C., Nayak, R., Jayswal, H. Computers
and Patel, A.
and Electronics 2021. Crop Recommendation
in Agriculture. 150 System
International Journal of Scientific & Technology (2018): 88–97.
Research. 8(10): 1121-1125. • Rajamanickam, J. and Savitha D.M. 2021. Predictive
• JuhiReshma, S.R., and D. John Aravindhar. model construction for prediction of soil fertility using
2021. Fertilizer estimation using deep learning decision tree machine learning algorithm." INFOCOMP
approach. Natural Volatiles & Essential Oils Journal. Journal of Computer Science. 20(1): 49-55.
5745-5752. • Reddy, A. A. (2019). The soil health card Scheme in
• Klompenburg, Thomas, Kassahun, Ayalew, Catal, India: Lessons learned and challenges for replication
Cagatay, 2020. Crop yield prediction using machine in other developing countries. Journal of Natural
learning: A systematic literature review. Computers Resources Policy Research, 9(2), 124-156.
and Electronics in Agriculture, Volume 177 • Sarker, I.H. 2021.Machine Learning: Algorithms, Real-
• Kotu, V. and Deshpande, B. 2019. Data Science World Applications and Research Directions. SN
Concepts and Practice (Second Edition). Morgan COMPUT. SCI. 2, 160.
Kaufmann Publishers Cambridge. p.548. • Savary, S., Willocquet, L., Pethybridge, S. J., Esker,
• Lal, R. (2015). Soil degradation as a reason for P., McRoberts, N., & Nelson, A. (2019). The global
inadequate human nutrition. Food Security, 7(5), 981- burden of pathogens and pests on major food crops.

86
 Nature Ecology &Evolution, 3(3), 430-439.
• Sharma, A., Jain, A., Gupta, P. and Vinay Chowdary,
V. 2020. Machine learning applications for precision
agriculture: A comprehensive review. IEEE Access.
9(2020): 4843-4873.
• Sharma, Neha, Sharma, Reecha, Jindal, Neeru,
2021.Machine Learning and Deep Learning
Applications-A Vision. Global Transitions
Proceedings,Volume 2, Issue 1.
• Singh, D. K., Singh, A. K., Singh, S. K., Singh, M. and
Srivastava, O. P. 2015. Effect of balanced nutrition
on yield and nutrient uptake of pea (Pisumstivum L.)
under Indo- Gangetic plains of India. The
Bioscan.10(3): 1245-1249.
• Smith, M. R., Reis Hodecker, B. E., Fuentes, D., &
Merchant, A. (2022). Investigating Nutrient Supply
Effects on Plant Growth and Seed Nutrient
Content in Common Bean. Plants (Basel,
Switzerland), 11(6), 737.
• Tscharntke, T., Clough, Y., Wanger, T. C., Jackson,
L., Motzke, I., Perfecto, I., Vandermeer, J. &
Whitbread, A. (2012). Global food security,
biodiversity conservation and the future of
agricultural intensification. Biological Conservation,
151(1), 53-59.
• Wan Yahya, Wan Mohd Nooriman & Abdullah, Abu
Hassan & Abdul Rahim, Norasmadi& Tan, Erdy.
(2021). Fuzzy logic based prediction of micronutrients
demand for harumanis mango growth cycles. Journal of
Physics: Conference Series. 2107

87