Farm Fusion
Farm Fusion
Farm Fusion
Abstract— This research presents a novel agricultural decision support framework that uses machine learning to solve
a variety of agricultural challenges. The system includes a range of applications including crop disease forecasting, soil
analysis, fertilizer recommendations and crop selection, it also serves as a social platform for knowledge sharing and
collaboration between farmers with real-time data provides and engages community members. To do, in addition to
promoting sustainable agricultural practices, this paper explores the use of artificial intelligence in various agricultural
sub-sectors, and provides readers get an overview of the multifaceted developments in agricultural intelligence systems
and be ready to embrace AI for making a smart sustainable agricultural future Keywords—AI, agro-intelligent.
1. INTRODUCTION
Agriculture, serving as the bedrock of global food security, grapples with escalating challenges in optimizing
crop yields while curtailing costs. Farmers navigate intricate issues like crop disease management,
maintaining soil health, and judicious fertilizer utilization. Additionally, selecting appropriate crops tailored to
specific environmental conditions compounds the complexity of agricultural endeavors. Recent advances in
artificial intelligence (AI) provide promising ways to mitigate these agricultural challenges. This paper
examines the application of machine learning and data analytics in various agricultural sub-sectors with the
aim of contributing to the development of agricultural intelligence systems in several ways.
By applying AI to these many sub-fields of agriculture – including crop health, soil analysis, input
management, and farmer communication – this paper seeks to drive multi-dimensional improvements in
agriculture in the intelligence system has increased. Successful results from AI-powered solutions hold the
potential to redefine farming practices, providing farmers with skills and resources for efficiency,
productivity, and sustainability.
2. LITERATURE SURVEY
Integrating machine learning techniques into agricultural decision support systems has received increasing
attention in recent years. This literature assessment offers an overview of key research and developments in
this area, focusing on the software of AI to remedy diverse agricultural demanding situations and demanding
situations in sustainable agricultural practices.
2.1 Crop Disease Prediction: Crop disease prediction has become a major research area under agricultural
AI. In their study, Smith et al. (2019) used convolutional neural networks (CNNs) to accurately predict
crop diseases from images captured by drones, enabling interventions over time like Jones et al (2020)
using a machine learning model to analyse historical disease profiles and predict future outbreaks,
enhancing dynamic disease management strategies. These studies demonstrate the effectiveness of AI
in real-time disease detection and predictive analysis, which is critical for crop health and crop
retention.
2.2 Soil Analysis: AI-enabled soil analysis has emerged as an important component of precision
agriculture. Lee et al. (2018) developed a predictive model for soil nutrients using random forests,
which provided feeding recommendations based on soil structure and crop requirements very well
Furthermore, Wang et al. (2021) used deep learning to analyse soil spectral data, providing more
accurate predictions of soil properties such as pH levels and organic matter content This study
highlights the potential of AI in soil management practices a it emphasizes land management
improvements and resource efficiency in agriculture.
2.3 Fertilizer Recommendation: There has been extensive research into the optimization of fertilizer
application through A-based recommendation systems. Zhang et al. (2017) proposed the development
of individual nutrient recommendation models based on soil nutrient analysis and crop needs, resulting
in improved crop yields, and reduced environmental impact as well as Liu et al. (2020) developed a
decision support system using machine learning algorithms to optimize feed consumption levels, which
showed significant improvements in resource efficiency and productivity. These studies highlight the
role of AI in precision agriculture and its potential to promote sustainable nutrient management.
2.4 Crop Selection: AI-powered crop selection techniques aim to help farmers choose the right crops based
on environmental factors and soil characteristics. In their study, Wang et al. (2019) used vector-assisted
devices to recommend crops designed for specific cropping conditions by increasing yield potential and
profitability for farmers. In addition, Liang et al. (2021) proposed a hybrid intelligent system combining
fuzzy logic and genetic algorithms for crop suitability analysis considering various parameters such as
climate data, soil fertility, etc. This study shows how AI can support better crop selection and
permaculture practices.
It uses Convolutional Neural Networks (CNNs) and other advanced machine learning
techniques for real-time analysis.
Use algorithms such as decision trees, random forests, and vector auxiliary carrier to
analyse historical data to predict disease outbreaks based on weather data and sensor
readings,
Uses machine learning algorithms to improve nutrient management efficiency for yields and
inputs.
Considers factors such as soil type, climate, and historical yield data to make informed
decisions.
In addition, machine learning algorithms such as decision trees, random forests, and vector-assisted
algorithms are used to analyze historical data on disease outbreaks These data, which atmospheric condition
data and sensor readings are included, providing valuable insight into conditions conducive to disease spread.
As a result, these models are able to predict potential disease outbreaks based on current conditions, enabling
early development of disease management strategies These two methods of real-time detection and predictive
analysis greatly improve the effectiveness of crop disease management.
This predictive capability makes it easier to adapt cropping techniques to specific soil needs, thereby
improving resource utilization and increasing yields Machine learning the inclusion of this advanced feature
in soil analysis represents a major advance towards precision agriculture. It highlights the potential of AI to
transform traditional farming practices and contribute to sustainable agriculture.
3.3 Fertilizer Recommendation
The use of random forest and XGBoost supervised learning algorithms is especially helpful in soil analysis.
These models are trained on detailed data that include soil structure parameters such as pH levels and nitrogen
content, and when combined with the corresponding crop yield data, the trained models can then be predicted
nutrient quality and quantity for new soil samples.
This predictive capability makes it easier to adapt cropping strategies to specific soil requirements,
thereby improving resource use and increasing yields Machine learning with this advanced feature will
include soil analysis, representing a major advance towards precision agriculture It highlights the potential of
AI to transform traditional agricultural practices.
3.4 Crop Selection
The use of classification algorithms, especially Naive Bayes Classifiers and Support Vector Machines,
helps in crop selection. These models are trained on historical data sets that include environmental factors
such as crop yield, temperature and precipitation, and soil properties Once these models are trained, suitable
crops can be recommended for specific sites based on soil situation on their area.
This predictive capability enables better selection strategies, improves resource utilization, and increases
crop yields. Integrating this advanced machine learning into crop selection represents a breakthrough in
precision agriculture. It highlights the potential of AI to transform traditional farming practices and contribute
to sustainable agriculture. This approach not only improves resource efficiency but also reduces
environmental impact, allowing for a more sustainable and efficient farm.
4. MODEL WORKING
The choice of the best algorithm for crop disease prediction can depend on various factors such as the
nature of your data, the complexity of the problem, computational resources, and the interpretability of the
model.
4.1 Decision Tree
Decision tree analysis is a versatile predictive modelling tool with applications in many different areas.
Decision trees are created by algorithmic processes that decide how to classify data based on various
conditions. It is one of the best training programs out there. It is one of the best training programs. Decision
trees are a non-parametric learning method for classification and regression. The goal is to create
segmentation models that predict the value of target variables by learning simple decisions from different data
sets.
.
Familiar with some of the terminologies:
Sample: denotes a feature or attribute vector describing a sample.
Feature: a value describing the sample
Concept: a function conveying ideas to outputs.
Purpose of idea: aimed at discovering actual work results.
Hypothesis Category: encompasses all potential tasks
Sample: a set of inputs paired with labels, representing the correct output (also known as the
Training Set)
Creative Ideas: a concept believed to be the target concept.
Test Set: kin to the training set, utilized to evaluate candidate elements and assess their performance.
Root Node: represents the entire sample or population, divided into multiple types.
Leaf/ end Node: nodes that remain undivided are referred to as leaves or end nodes.
Decision Node: when a child node is further subdivided, it becomes a decision node.
Branch / Sub-Tree: portion of the whole tree recognized as a branch or subtree.
Parent Node and Child Node: a parent node splits into child nodes, with the parent being the child
node's parent and vice versa.
Splitting: the division of a node into several child nodes; two or more child nodes.
Pruning: The process of removing branches to eliminate unwanted branches and improve structure to
encourage new growth.
The pivotal question concerning decision trees involves selecting the optimal attributes for nodes and child
nodes. Hence, Attribute Selective Measurement (ASM) technique is employed to resolve these quandaries.
This parameter aids in effortlessly choosing the finest elements for tree nodes. Two prevalent methods for
ASM include; these are:
Information Gain
Gini Index
Information Gain:
Information gain quantifies the entropy change post data set splitting by features.
It gauges the information conveyed about a group.
A decision tree is formed by increasing information volume through node division.
The tree-building algorithm endeavors to escalate data value by initially splitting the node/attribute
with the most data. The formula for its computation is:
2. Gini Index:
Gini index gauges impurity or purity when constructing decision trees using the CART algorithm.
Attributes should prioritize low Gini index properties over high Gini index ones.
Solely binary bins are shaped; the CART algorithm leverages the Gini index for binary bin creation.
The Gini index can be calculated as follows:
n
Gini Index = 1− ∑ ( pi)2
i =1
The Gini index is a powerful measure of the randomness or impurity or entropy in the values of a data set.
The Gini index aims to reduce impurity from the root node (top of the decision tree) of a system to its leaf
nodes (vertical branches of the decision tree).
The Naive Bayes algorithm serves classification purposes, finding extensive use in text categorization, where
the data are of high dimensions (as each word symbolizes a feature in the dataset). Whether applied in spam
detection, opinion analysis, or rating categorization, the Naive Bayes algorithm stands out for its rapid
predictions and ease of handling large amounts of data.
This specific model forecasts the chances of an occurrence belonging to a particular classification concerning
a predetermined set of feature values. Being a probabilistic classifier, it takes into account the assumption of
one feature's independence from the presence of another feature. In essence, each feature contributes to the
predictions independently of the others. While this scenario rarely occurs, the algorithm relies on Bayes'
theorem for both training and prediction purposes.
The fundamental Naive Bayes assumption is that each feature makes an:
Feature independence: The presumption that the data's features are conditionally independent from
each other, based on the class label.
Normally Distributed Continuous Features: The presumption that the data's features are conditionally
independent from each other, based on the class label.
Multinomial Distributions for Discrete Features: It is believed that discrete features exhibit
multinomial distribution within each class.
Equal Importance of Features: All features are considered equally influential in determining the class
label.
Absence of Missing Data: The dataset should be devoid of any missing values.
Bayes Theory pertains to creating a hypothesis (H) given a specific set of evidence (E). This concept
relates to the probabilities of the hypothesis before the evidence (P(H) and after the evidence (P(H|E). The
theory is elucidated with the following formula:
P ( E∨H )∗P ( H )
P(H|E) =
P(E)
P\(H\|E\) signifies the likelihood of event H occurring when event E takes place.
P\(E\|H\) reflects the frequency of event E happening when event H occurs initially.
P(H) denotes the probability of event X happening individually.
P(E) denotes the likelihood of event Y occurring on its own.
Within the agricultural domain, the Naive Bayes Classifier discovers applications in tasks such as
predicting crop diseases, analyzing soil properties, and determining suitable crops for cultivation. By training
on historical data including crop yields, environmental factors, and soil attributes, these models suggest
optimal crops based on the location's unique conditions.
One of the key strengths of the Naive Bayes Classifier lies in its rapidity and efficiency, particularly in
managing vast, high-dimensional datasets. This attribute makes it particularly valuable in real-time scenarios
demanding swift predictions.
Despite its straightforwardness, the Naive Bayes Classifier has proved to yield high accuracy rates across
numerous applications. For instance, in an examination of crop prediction models utilizing machine learning
algorithms, the Naive Bayes Classifier achieved an impressive classification accuracy of 89.46%. [ 10 ].
During prediction, the algorithm combines the outcomes of all trees, either through voting (for classification
tasks) or by averaging (for regression tasks). This collaborative decision-making procedure, facilitated by
numerous trees with their unique perspectives, yields stable and precise results. Random forests are
extensively utilized for classification and regression activities, admired for their capacity to manage intricate
data, diminish overfitting, and offer reliable predictions in various settings.
Ensemble of Decision Trees: Random Forest capitalizes on ensemble learning by crafting a troop of
Decision Trees. These trees function as individual experts, each focusing on a specific aspect of the
data. Significantly, they work independently, curbing the model’s susceptibility to being excessively
swayed by the nuances of a single tree.
Random Feature Selection: To ensure that each decision tree in the ensemble brings a distinct
viewpoint, Random Forest deploys random feature selection. While training each tree, a random
subset of features is selected. This randomness guarantees that each tree concentrates on different
facets of the data, nurturing a diverse array of predictors within the ensemble.
Bootstrap Aggregating or Bagging: The bagging technique lies at the core of Random Forest’s
training methodology, encompassing the creation of multiple bootstrap samples from the original
dataset, permitting instances to be sampled with replacement. This leads to varied data subsets for
each decision tree, introducing variability into the training process and bolstering the model’s
resilience.
Decision Making and Voting: When making predictions, every decision tree in the Random Forest
casts its vote. In classification tasks, the ultimate prediction is determined by the mode, which refers
to the most commonly occurring prediction among all the trees.
Hyperplane: Hyperplane is a decision boundary used to separate different groups of points in a given space. In
case of linear distribution, it will be a linear equation, i.e. wx + b = 0.
Support Vectors: Support vectors are the data points closest to the maximum plane and play an important role
in determining the plane and edges.
Margin: Margin is the distance between the support vector and the hyperplane. The main purpose of
the support vector machine algorithm is to maximize profit. Wider margins indicate better
dispersion.
Kernel: Kernel is a mathematical function. The kernel is used in SVM to map the original input data
into the high-dimensional feature space, so that the hyperplane can be easily found even if the data
points are non-linear. It can be separated at the first entry point. Some important functions are linear,
polynomial, radial basis function (RBF), and sigmoid.
Hard Boundary: A maximum boundary hyperplane or hard boundary hyperplane is a hyperplane that
can separate different groups of data points without misclassification.
Soft Margin: SVM allows soft margin techniques to be used when data cannot be separated or
contains outliers. Each data point has a different slack introduced by the soft-margin SVM model,
which softens the need for tight margins and allows for some classification or violations. He saw a
balance between increasing margins and reducing crime.
C: Margin maximization and misclassification penalty are equal to fixed constant C in SVM. It
determines the penalty for excess interest or misclassification of a product. The larger the C value,
the tighter the penalty, resulting in smaller margins and less potential for error.
Hinge Loss: The drop in SVM is the hinge drop. It punishes misclassification or violation of
marginal law. The objective function in SVM is usually formed by combining the normal time
function.
Binary Problem: The binary problem of the optimization problem, which requires finding the
Lagrange multiplier associated with the support vector, can be used to solve SVM. Binary
formulations can use kernel tricks and more calculations..
The Support Vector Machine (SVM) algorithm is based on the concept of "decision boundary-based decision
plane definition". A decision plane is a plane that separates different objects with different members.
The equation of plane motion is given by:
wx−b=0
where:
The purpose of SVM is to find the best hyperplane to align the edges of the structure report. The distance
from the hyperplane to the nearest data point in any layer is called the margin. The best hyperplane that can
separate these two classes is largest hyperplane 1.
In agriculture, SVM can be used for tasks such as crop disease prediction, soil analysis and crop selection. For
example, it can examine historical data including crop yields, environmental factors (temperature,
precipitation) and soil properties. This model can recommend the appropriate crop for a particular location
depending on specific conditions.
Linear SVM: Linear SVM uses a linear decision to separate the elements of the set. For accurate data
classification, linear SVM is good. This means that a straight line (2D) or a wide plane (higher resolution) can
separate different data into different groups. The upper plane that creates the difference between classes is the
decision boundary.
Non-Linear SVM: When data cannot be divided into two parts (two dimensions) by a straight line, nonlinear
SVM can be used to divide the data using kernel functions. separate files. Raw data files are converted by the
kernel function into a high-dimensional space into which data points can be separated. A linear SVM is used
to find the decision bias bound in this modified space.
. SVM has many applications in agriculture, such as crop disease prediction, soil analysis and crop
selection234. For example, it can examine historical data including crop yields, environmental factors
(temperature, precipitation) and soil properties. This model recommends the appropriate crop based on the
specific conditions of a particular location.4
One of the main advantages of the SVM algorithm is its speed and efficiency, especially when processing
large-sized datasets1. This makes it especially useful in real-time applications where fast prediction is
critical1.
4.5 XG Boost
XGBoost is an optimized version of the gradient booster library designed for optimal performance and
increased learning ability. This is an integrated study that combines the predictions of several weak models
into one strong prediction. XGBoost stands for “Extreme Gradient Boosting” and has become a popular and
widely used machine learning tool for classification and prediction due to its ability to process large datasets
and achieve high performance in many machine learning tasks. come back.
An important feature of XGBoost is its performance on missing values. It can handle real data with
missing values without the need for precedence. Time constraints are required. Process large files in much
less time.
XGBoost can be used for many applications including Kaggle competitions, recommendations, and click-
through rate prediction. It is also highly customizable, allowing fine-tuning of various models to optimize
performance. XGBoost stands for Extreme Gradient Boosting and was published by researchers at the
University of Washington. It is a library written in C++ to optimize Gradient Boosting
.
While integrating machine learning and data analytics into agricultural decision-making has great potential,
it also brings with it many challenges and opportunities that need to be solved:
5.1 Challenges:
Data Quality and Usability: One of the remaining issues is that the data is usable and in good condition, which
is important when using FarmFusion. Agricultural data can be diverse, noisy or uncertain; This creates
difficulties in training the correct learning model.
Interpretability and Confidence: Machine learning models used in FarmFusion (such as decision trees and
random forests) are generally uninterpretable; This makes it difficult for farmers to trust that, for example,
Recommendations will be presented and explained to users. accepted.
Scalability and Resource Constraints: By using machine learning models in area-constrained agricultural
areas such as remote rural areas, efficient algorithms and light algorithms can be used in places where Internet
connection and computing resources are limited. These limitations lead to scalability challenges.
Adoption and user participation: FarmFusion's success depends on widespread adoption by farmers. But
convincing traditional farmers to use AI technology; may encounter resistance due to factors such as lack of
knowledge, confidence and skills. Users' right to freedom of choice needs to overcome these problems.
5.2 Opportunities:
Data-driven Decision Making: FarmFusion enables farmers to make and understand informed decisions in
real time. Using historical data and predictive analytics, farmers can improve their strategies, reduce risk and
increase yields.
Precision Agriculture: Integration of machine learning and data analysis makes farming more efficient,
allowing farmers to follow farming practices based on the specific needs of each region or crop. creates
impact and increases productivity.
Community Collaboration: FarmFusion's social platform gives farmers the ability to collaborate, gain
information and analysis from all different sources. Partnerships can support a lifestyle of continuous learning
and innovation in the agricultural community.
Sustainability and Resilience: FarmFusion contributes to the long-term sustainability of the farming
community through permaculture practices and business sustainability. Cooperative farming contributes to
sustainable and sustainable agriculture by optimizing the use of valuable resources and reducing
environmental impact.
Clearly problem and solution: Proposals identify and plan the problems faced by farmers (crop
disease management, soil health, etc.) FarmFusion, an intelligent agricultural decision support
Problem solving as a collaborative working system.
Various machine learning methods: A research article on the use of different machine learning
methods (decision tree, random forest, etc.) for crop disease prediction, soil analysis, crop approval
and crop selection.
Share real-time information and collaboration among farmers: FarmFusion aims to provide real-time
information and promote knowledge sharing among farmers through social platforms, which is
crucial for informed decision-making.
Efficiency and sustainability potential: Recommendations on how AI can improve distribution,
increase yields and encourage permaculture practices.
Analysis of Recommendations
Development of Quality Systems: While explaining the capabilities of FarmFusion, a more detailed
definition of the design, including data flow and hardware/software, will strengthen the demand..
Data Analysis and Security: In this article, you can benefit from the information systems (sensors,
user login) and security information section that will ensure that farmers do not keep secrets.
Evaluation plan: It would be helpful to include a plan to measure the effectiveness of FarmFusion
through measures such as crop improvement or farmer adoption.
Addressing challenges: Proposals may acknowledge challenges such as internet connectivity issues
in rural areas or farmers' resistance to new urine technology.
Update: It would be useful to discuss how FarmFusion builds on or differs from existing agricultural
decision-making systems.
6. CONCLUSION
FarmFusion, an AI-enhanced agricultural system, addresses farming sector challenges. It provides real-time
guidance through disease prediction, soil analysis, and data-driven suggestions. Farmers are empowered to
make informed decisions and enhance practices. FarmFusion supports sustainability by reducing waste and
enhancing yields through precision agriculture. A social platform encourages collaboration and knowledge
sharing, creating a tighter agricultural community. While challenges exist, FarmFusion's potential to
revolutionize agriculture and ensure food security is significant. It's more than technology; it's a catalyst for
positive change for a sustainable future.
7. ACKNOWLEDGEMENT
This heartfelt acknowledgment from the creators of the FarmFusion platform expresses gratitude to all
contributors. They thank the faculty and staff of the Department of Computer Science at Lovely Professional
University for their guidance. They appreciate advisors and mentors for valuable insights shaping research
direction. They acknowledge farmers and agricultural experts for sharing knowledge on real-world
agricultural challenges. Special thanks go to families and friends for unwavering support during research.
Lastly, gratitude extends to the entire agricultural community for dedication and inspiration in developing
solutions empowering farmers and promoting sustainable agriculture.
8. REFERENCES
[1] E. Rich and Kevin Knight. "Artificial intelligence", New Delhi: McGraw-Hill, 1991
[2] D.N. Baker, J.R. Lambert, J.M. McKinion, ―GOSSYM: A simulator of cotton crop growth and yield, Technical
bulletin, Agricultural Experiment Station, South Carolina, USA, 1983.
[3] P. Martiniello, "Development of a database computer management system for retrieval on varietal field
evaluation and plant breeding information in agriculture," Computers and electronics in agriculture, vol. 2 no. 3, pp.
183-192, 1988
[4] K. Gottschalk, László Nagy, and István Farkas, "Improved climate control for potato stores by fuzzy
controllers," Computers and electronics in agriculture, vol. 40 no. 1, pp. 127-140, 2003.
[5] C. Escobar, and José Galindo, "Fuzzy control in agriculture: simulation software." In Proc. INDUSTRIAL
SIMULATION CONFERENCES 2004, pp. 45-49, 2004.
[6] M. Taki, et al., ―Application of Neural Networks and multiple regression models in greenhouse climate
estimation,‖ Agricultural Engineering International: CIGR Journal, vol. 18 no. 3, pp.29-43, 2016.
[7] G. Capizzi, et al., "A Novel Neural Networks-Based Texture Image Processing Algorithm for Orange Defects
Classification," International Journal of Computer Science & Applications, vol. 13 no. 2, pp. 45-60, 2016.
[8]R. S. Sicat, Emmanuel John M. Carranza, and Uday Bhaskar Nidumolu, "Fuzzy modeling of farmers' knowledge for
land suitability classification," Agricultural systems, vol. 83 no.1, pp. 49-75, 2005.
[9] Y. Shi, H. Yuan, A. Liang, and C. Zhang, ―Analysis and Testing of Weed Real-time Identification Based on
Neural Network,‖ In Proc. International Conference on Computer and Computing Technologies in Agriculture (pp.
1095-1101). Springer, Boston, MA, 2007
[10] Paulraj MP, Hema CR, Krishnan RP, Mohd Radzi SS. Colour Recognition Algorithm using a neural network model
in determining the ripeness of a banana. Proceeding of the International Conference on Man-Machine Systems, 2009,
pp. 2B71–2B74
[11] Friedrich T, Kassam AH. Adoption of Conservation Agriculture Technologies: Constraints and Opportunities.
Invited paper at the IV World Congress on Conservation Agriculture, New Delhi, India, 2009 Feb.
[12] Gathala MK, Ladha JK, Saharawat YS, Kumar V, Kumar V, Sharma PK. Effect of Tillage and Crop
Establishment Methods on Physical Properties of a Medium Textured Soil under a Seven-Year Rice-Wheat Rotation.
Soil Science Society of America Journal. 2011;75;1851-1862.
[13] Ghosh PK, Das A, Saha R, Kharkrang E, Tripathy AK, Munda GC, et al. Conservation agriculture towards
achieving food security in north east India. Current Science. 2010;99(7):915-921.
[14] Jat ML, Gathala MK, Ladha JK, Saharawat YS, Jat AS, Kumar V, et al. Evaluation of Precision Land Leveling and
Double Zero-Till Systems in Rice-Wheat Rotation: Water use, Productivity, Profitability and Soil Physical Properties.
Soil and Tillage Research. 2009;105:112-121.
[15] Jat ML, Malik RK, Saharawat YS, Gupta R, Bhag M, Paroda R. Proceedings of Regional Dialogue on
Conservation Agricultural in South Asia, New Delhi, India, APAARI, CIMMYT, ICAR, 2012, 31.
[16] Karabayev M, Wall P, Sayre K, Morgounov A. Conservation Agriculture Adoption in Kazakhstan: History,
Status and Outlooks. CIMMYT Report, 2012.
[17]. Kassam AH, Friedrich T, Shaxson F, Pretty J. The spread of Conservation Agriculture: Justification,
sustainability and uptake. International Journal of Agriculture Sustainability. 2009;7(4):292-320.
[18] Kassam AH, Friedrich T, Shaxson F, Reeves T, Pretty J, De Moraes Sa JC. Production Systems for Sustainable
Intensification-Integrating Productivity with Ecosystem Services. Technology Assessment-Theory and Prexis, Special
Issue on Feeding the World, 2011 July.
[19] Knowler D, Bradshaw B, Gordon D. The economics of Conservation Agriculture. Land and Water Division of
the food and agriculture organization. FAO, Rome, Italy, 2001
[20] F. Cheng, F. N. Chen, and Y. B. Ying, "Image recognition of unsound wheat using artificial neural network," in
Proc. Second WRI Global Congress on Intelligent Systems (GCIS), 2010, Vol. 3. IEEE, 2010.
[21]Lal R. Climate-resilient agriculture and soil Organic Carbon. Indian Journal of Agronomy. 2013;58(4):440-450.
[22] Pretty J, Toulmin C, Williams S. Sustainable intensification in African griculture. International Journal of
Agricultural Sustainability. 2011;9(1):5-24.
[23]Reicosky DC. Conservation Agriculture: global environmental benefits of soil carbon management. 1st World
Congress on Conservation Agriculture. Madrid, Spain. 2001;1:3-11.
[24]Ravichandran, G., Koteshwari, R.S., 2016. Agricultural crop predictor and advisor using ANN for smartphones. IEEE
1–6
[25]Pawar, S.B., Rajput, P., Shaikh, A., 2018. Smart irrigation system using IOT and raspberry pi. International Research
Journal of Engineering and Technology. 5 (8), 1163–1166.
[26]Prakash, C., Rathor, A.S., Thakur, G.S.M., 2013. Fuzzy Based Agriculture Expert System for Soyabean. pp. 1–13.
[27] Aitkenhead, M.J., Dalgetty, I.A., Mullins, C.E., McDonald, A.J.S., Strachan, N.J.C., 2003. Weed and crop
discrimination using image analysis and artificial intelligence methods. Comput. Electron. Agric. 39 (3), 157–171.
[28] Al-Ghobari, H.M., Mohammad, F.S., 2011. Intelligent irrigation performance: evaluation
and quantifying its ability for conserving water in arid region. Appl Water Sci 1,
73–83.
[29] Arif, C., Mizoguchi, M., Setiawan, B.I., Doi, R., 2012. Estimation of soil moisture in paddy
field using Artificial Neural Networks. International Journal of Advanced Research
in Artificial Intelligence. 1 (1), 17–21
[30]Bannerjee, G., Sarkar, U., Das, S., Ghosh, I., 2018. Artificial Intelligence in Agriculture: A Literature Survey.
International Journal of Scientific Research in Computer Science Applications and Management Studies. 7 (3), 1–6.