Crop Yield Prediction Using Machine Learning

Muthamizh Selvam Shaik Muskan Darmaraju Deepika

Asst. Professor Dept. of AIML Student, Dept. of CSE Student, Dept. of CSE
QIS College of Engineering And Qis College of Engineering and QIS College Of Engineering and
Technology ,A.P,India Technology, A.P, India Technology,A.P ,India
[email protected] [email protected] [email protected]

Sanikala Lekhana Madala Jaswanth Kommanaboyina Yeshwanth Yadav

Student, Dept. of CSE Student, Dept. of CSE Student, Dept. of CSE
QIS College Of Engineering and QIS College Of Engineering and QIS College Of Engineering and
Technology,A.P ,India Technology,A.P ,India Technology,A.P ,India
[email protected] [email protected] [email protected]

Vamaraju Naga Sai Sahithi

Student, Dept. of CSE
Qis College of Engineering and
Technology, A.P, India
[email protected]
Abstract: be receiving information about the state where the edit is
produced. Season and year play a very important role in
Crop yield prediction is a critical aspect of agricultural making these predictions practical and custom fitted. Indeed,
management, aiming to optimize production and resource it determines the time of the year when crops are planted-at
allocation. This study leverages machine learning techniques dry season, rainstorm or winter-which directly affects
to develop a predictive model for crop yields, utilizing development and abdicate results.
historical data on weather conditions, soil properties, rainfall
and crop management practices. By employing algorithms Therefore, involvement of this year enables the stage to
such as random forest, linear regression, decision trees, and make use of the verifiable trends and patterns like freakish
support vector machine, we aim to achieve high prediction weather or perhaps climatic changes for enhancing the
accuracy. credence of the predictions. This holistic approach with
which the expectation given up does not depend on the single
The model is trained and validated on extensive datasets, variable but a composite condition of the real world make
with performance evaluated using metrics like mean absolute that makes the crop more unsurprising and effective. The
error and root mean square error. Our approach offers a data- Edit Surrender Expectation Venture provides farming people
driven solution to enhance decision-making in agriculture, with significant amounts of knowledge of how to handle all
potentially leading to improved crop productivity those variables into one easily deployable stage. It will make
and sustainability. a difference for them to prepare beforehand, manage their
assets better, and take smarter decisions which can lead them
Future engineering techniques are employed to preprocess
to advance efficiency and profits. This stage more than
the data and extract meaningful information, followed by
reasonably enough predicts yields, serves as a basis for
model training and evaluation.
economical practices in agriculture, and helps the farmer
Keywords-: adapt to a changing climate and secure better crops for yet
another year. It would be able to contribute to a more
Crop Yield Prediction, Agricultural Management, Machine
adaptive and forward-looking agrarian sector.
Learning, Weather Conditions, Soil Properties, Rainfall,
Crop Management Practices, Random Forest, Linear II. Literature review
Regression, Prediction Accuracy, Mean Absolute Error, Root
Mean Square Error, Data-Driven Agriculture, Crop Much literature survey is made regarding the various
Productivity, Sustainability, Feature Engineering, Model approaches that researchers have used to predict crop yields
Training, Model Evaluation. with the aid of machine learning techniques for the paper
"Crop Yield Prediction Using Machine Learning." Various
studies have been shown to provide accurate yield
I. Introduction predictions by algorithms, like random forests and neural
networks and the strategy of support vector machines.
The Trim Abdicate Forecast Venture is particularly Recently, for instance, Wang et al. (2019) applied random
designed to help empower such farmers and agricultural
forests to make predictions of maize yields in China; these
professionals with a smart, data-driven tool for predicting
performed incredibly well compared to older, more
on-time yields based on a spectrum of such basic factors.
Cultivation is always unpredictable-the yields are always on established ways of accomplishing similar tasks. Another
the move with changing natural conditions as well as the application of neural networks was the prediction of rice
resources available. This extend makes a difference lessen yields in India, a problem addressed by Mondal (2020).[3]
vulnerability by checking out basic parameters, for example, These models can perhaps address complex variables such
precipitation, zone, trim sort, fertilizers, pesticides, state,
as weather patterns and satellite imagery to give reliable
season, and year to give more specific and reliable
results. Other studies, such as Gupta et al. (2021), used deep
expectations. Doing this makes a difference ranchers make
way better choices around trim administration and long-term learning models to predict cotton yields. Such studies are
arranging. One of the most basic factors that affect the yields therefore indicative of the strength of machine learning in
is precipitation-available water for the plant growth. helping farmers. Overall, these studies conclude that,
compared to traditional methods, techniques based on
Designs of rainfall are followed by the stage so that available machine learning are better and can be an extremely
water is taken into accounts of the surrender estimates. The valuable resource for helping farmers make even more
region under development estimate is very important to effective use of available resources and maximize yields.
determine. The sort of edit being planted also is similarly
Still, much work remains to be done in refining these
important because each trim reacts differently toward the
models so that they can be applied practically in actual
climate, soil, and the style of farming, making crop-specific
estimates more rapid for the farmer. agricultural situations.[5]

Directly the use of fertilizers and pesticides has an impact Editing surrender forecast using machine learning in a
on the quality of an edit concerning growth and security from writing survey finds advancements from the very basic
pests and diseases. Fertilizers enhance plant health, as they process of machines-the calculations such as SVMs and
provide essential nutrients. The use of pesticides prevents or BPNNs-to more progressed versions like CNNs and XGB.
reduces the damage caused by pests. Further, the degree of Early datasets showed little datasets and destitute execution;
regional variation is taken into account because the users will
however, by utilizing cutting-edge strategies, bigger Machine Learning Approaches: The proposed machine
datasets, and complex designs, altogether more precise learning approach aims to predict the best crop yields by
forecasts came. The audit underlines that the natural parts, analyzing various atmospheric factors (such as rainfall,
for example, soil, climate, and trim administration, together temperature, and humidity) and land factors (including soil
with information preprocessing techniques, are important to pH and soil type). This system is designed to assist farmers
improve demonstrate proficiency. These innovations in in making informed decisions about crop selection based on
machine learning result in economical and information- the collected dataset.
driven agrarian hones [9]. The literature review in "A
Comprehensive Review on Machine Learning Approaches Data Mining Techniques: The crop yield prediction process
for Yield Prediction Using Essential Soil Nutrients" utilizes a Data Mining approach to analyze soil datasets,
discusses various machine learning techniques and their which is essential for improving the accuracy of yield
applications in predicting agricultural yield based on soil predictions [6].
nutrient data[8]. The review highlights the significance of Support Systems for Farmers: There are existing support
soil health and nutrient availability in determining crop systems, such as one in Bangladesh, that aim to assist poor
yield, emphasizing that the quality of crop output is closely farmers with crop demand through a website. However,
linked to soil fertility [2]. challenges arise as uneducated farmers struggle to use these
The authors reference multiple studies that have explored systems effectively.
different machine learning models, such as Support Vector Overall, the literature emphasizes the need for effective crop
Machines (SVM), Extreme Learning Machines (ELM), and yield prediction systems to improve agricultural practices
deep neural networks, for predicting soil fertility and crop and support farmers in making better decisions.
yield. The review also notes the importance of precision
agriculture (PA) in developing intelligent support systems III. Proposed System & Architecture
that consider various crop parameters, including soil type,
nutrient presence, and environmental factors [4]. 1. Input Layer (Information Collection)
Firstly, the information collection from a range of variables
The paper "Crop Yield Prediction using Machine Learning
that influence crop yield begins.The information inputs
and Deep Learning Techniques" explores how modern
include:
algorithms might help farmers in the region of Rajasthan, Weather information: Temperature, rainfall, humidity, etc.
India to better predict crop yields. The researchers have Historical crop yield data: Past yields for particular crops at
tested numerous methods to forecast yields for key crops different locations.
like wheat, mustard, and barley, including some most used Satellite and sensor data: Crop health, growth, etc. (NDVI
such as Random Forest, SVM, Gradient Descent, LSTM, measure from a satellite)
Lasso regression. Of these, the Random Forest model These information inputs are detected from the ground by
showed to give the highest-accuracy model at 96.3% sensors or climate stations, or detected from space by means
prediction accuracy. The study made use of data spanning of inaccessible detecting innovation like satellites or
1997-2019 with elements such as soil and rainfall for rambles, and are held in a central database or in cloud
making more stable predictions. The paper connects crop storage.
data with environmental factors, which shows how 2. Information Preprocessing Layer
technological means can strengthen farmers' ability in
making wiser decisions as to what crops they are supposed After collecting the information, it is cleaned and arranged
to plant and how resources should be implemented for the for utilize inside machine learning models:
efficiency and betterment of agricultural outcome.[4] Cleaning: Removing lost or off base data.
Normalization: Scaling so that all highlights can be
The report "Optimizing Trim Yields through Machine compared on the same scale, hence all being rise to for a
Learning-Based Expectation" shows how machine learning highlight like temperature and soil pH.
is changing farming. It talks about algorithms like Highlight Engineering: Creating modern highlights, such as
Credulous Bayes, Arbitrary Timberland, and Bolster Vector normal rain in a month.
Machines. These helps predict edit yields by looking at Data Part: Part the information into a preparing dataset and a
things like temperature, stickiness, rain, and soil additives. testing dataset, permitting for preparing of the show and
The study also looks at ways to make predictions more ensuing testing of the last mentioned in terms of its accuracy
in predicting.
accurate such as Recursive Include Disposal and Engineered
Minority Oversampling. This helps farmers make better 3. Include Choice Layer: Identify the most critical
choices and grow more crops [1]. Importance of highlights of the information as dependable for trim
Agriculture: Agriculture plays a crucial role in developing abdicate. These include:
regions like India, where a significant portion of the Correlation Investigation: Appears the distinctive factors-
population relies on farming for their livelihood. However, that are temperatures or rain-that have an affect on yield
its contribution to the GDP is relatively low, primarily due Recursive Include End (RFE): this is a highlight disposal
to inadequate decision-making regarding yield prediction. calculation that points at expelling the slightest mindful
highlights for portraying the issue and held the most Model Retraining: The system will automatically retrain a
extreme mindful highlights to the expectation errand at model based on the latest information to ensure that the
hand. expectations are precise as conditions change around it.
4. Modeling Layer (Algorithms) IV. Methodology
There, we utilize three machine learning calculations to
make prescient models as appeared below: Data Collection Data Preprocessing
Linear Regression: How it works: This is one of the easiest
models whereby it expect a direct relationship between the
input highlights, such as temperature and precipitation, and Model Selection Model Training
edit surrender. The demonstrate employments an endeavor
to fit a line to best anticipate the yield given these features.
Use case: Reasonable for basic edit abdicate expectation
when the relationship between the factors is primarily linear. Deployment Monitoring
XGBoost: How it works: This demonstrate is more
progressed. It works by building **choice trees** step by
step. XGBoost is an outfit strategy. That implies, in
quintessence, it combines the yield of numerous littler,
numerous- tree models into a more grounded expectation
show. That makes it able of capturing much more complex
variable connections, such as temperature and precipitation
together influencing yield.
Application: XGBoost is valuable when dealing with high-
dimensional data having many features influencing the
abdicate of crops in a nonlinear way.
LSTM: How it works: It is a type of recurrent neural arrange
(RNN) utilized for the reason of time-series information.
LSTM especially does a brilliant work at learning designs in
changing information that are concerning time arrangement
such as crop yield. LSTM, in reality, examines
arrangements of information, like climate designs over the
growing season, and learns how those designs affect the The technique for the Trim Surrender Forecast Location
yield. incorporates a few key steps to ensure correct and strong
Use case: LSTM is helpful when past conditions (e.g., trim surrender forecasts based on parameters like
rainfall from previous months) are necessary for predicting precipitation, run, trim sort, fertilizers, pesticides, state,
the current crop yields. season, and year. The strategy begins with data collection
5. Evaluation Layer: Evaluation After training the models, from distinctive sources such as government agrarian
they are evaluated to measure how well they can predict records, meteorological databases, and open-source stages
crop yield. To test which model can better predict crop yield like FAO and NASA. This data consolidates climate plans,
on inconspicuous information, the model is tried by using rustic sharpens, and regional characteristics critical to alter
the **test dataset**. Common measurements of evaluation yields. Once collected, the data is cleaned and preprocessed
are: to handle misplaced values and abnormalities, while
categorical components such as trim sort and season are
Mean Squared Mistake (MSE): Average squared contrast encoded into numerical bunches utilizing strategies like one-
between the real versus predicted yields. Lower is better. hot encoding. Nonstop variables like precipitation and zone
R-squared (R²): Calculates how well change in crop yield is are scaled for appear optimization.
captured by the model.
6. Model Deployment Layer: Once the model performs Feature building is associated to make interaction terms
well, it gets sent to make real-time predictions: between components, such as precipitation and season or
fertilizers and alter sort, to capture complex associations that
API or Web Application: Users, either farmers or affect trim yield. The expand utilizes two machine learning
agricultural specialists, may input information around the models:Straight Backslide and XGBoost. Straight Backslide
ranchers, such as climate or soil, with a friendly interface, or will serve as a standard illustrate, giving a essential
through API calls, to make expectations around crop yields. relationship between the input variables and alter yield. In
7. Observing and Upkeep Layer: It is employed in separate, XGBoost, a more advanced gathering learning
ensuring that the model continues working well over time. calculation, will handle non-linear associations and
shrewdly, moving forward the prediction's precision by
Performance Checking: Checks the accuracy of the forecast progressively building choice trees. Hyperparameters such
at given time intervals and updates the model as more data as learning rate, tree significance, and estimators will be
becomes available. tuned to avoid overfitting.
The models will be arranged utilizing a portion dataset (80% application for other crops, taking into account different
planning and 20% testing), with execution surveyed climatic and economic factors.
utilizing estimations such as Pitiless By and large Botch
Versatility and Client Encounter: The site needs to adapt to
(MAE), Pitiless Squared Botch (MSE), and the R-squared
manage large numbers of clients during busy times such as
score. The illustrate with the finest prescient execution will
planting or harvest seasons. Cloud-based structures (e.g.
be chosen for course of action on the location. The location
AWS, Google Cloud) provide solutions to scale and handle
will be built utilizing front-end progresses like HTML, CSS,
big datasets. An easy-to-use friendly interface is key.
and JavaScript to supply a user-friendly interface. The back-
Because the end users might be farmers with different levels
end will be made in Python utilizing frameworks such as
of tech know-how, the site should work well on mobile
Carafe or Django to have the machine learning appear.
devices with simple input areas and visual aids like charts or
Clients will input their parameters, and the illustrate will get
maps to help understand predictions more easily.
ready the data to supply real-time trim resign expectations.
Real-time and Dynamic Data Integration: Integrating up-to-
At final, the appear will be sent on cloud stages like AWS or
the-minute weather data (through APIs from weather
Google Cloud for flexibility and accessibility, with standard
services) will allow the system to give more current and
overhauls based on cutting edge data to ensure nonstop
exact forecasts for rainfall and pest outbreaks. The system
illustrate exactness. This comprehensive procedure planning
can also make pesticide and fertilizer suggestions based on
advanced machine learning techniques with an accessible
current weather conditions.
web organize, engaging agriculturists and agrarian
accomplices to frame taught, data-driven choices to VI. Conclusion
optimize edit efficiency.
The conclusion is that even though there are significant
V. Discussion benefits in designing an edit surrender forecast site with
climatological, geographical parameters, range of fertilizers,
Information Collection and Integration: This extend
pesticides, state title, season, year, and trim sort parameter
highlights the collection of high quality, precise information
there are other aspects of such development which need
as its foundation. This includes historical information and
addressing. The success of such a platform will depend on
up to the minute information relating to rainfall, fertilizer
quality of data collected, high end machine learning
and pesticide application, area under cultivation and crop
techniques and the consideration of spatial and temporal
types. Information needs to be gathered on the basis of the
variation. Linking instant weather data and ensuring Sling is
region meaning the information has to be collected on a
essential to accurate and timely forecasts.
state or even a local sector. Government and open data
sources for example IMD (Indian Meteorological Office) Equally important is the usability of the platform with
and ICAR can be tapped. emphasis on designing a user friendly and mobile
appropriate user interface for farmers and other users. With
Dealing with Complex Factors: There are numerous
the proper approach, this website can help by supporting the
interactions between, for example, precipitation, pesticide
farmers in making decisions based on factual data and hence
application or fertilizer use. On the other hand, whether
improving crop management and yield outcome, which in
these fertilizers are efficient or not may also greatly depend
the end benefits agricultural productivity.
on rainfall, the type of soil and when it is applied. Advanced
machine learning models like Random Forest, Gradient VII. References
Boost or Neural Nets are appropriate to model such
dependencies. However, it is important to balance model 1. Harshith, M., Sahu, A., Indrakanti, S., Reddy, R. K., &
complexity and interpretability in order to help users Bhutada, S. (2023). Optimizing Crop Yields through
understand why a certain prediction has been made. Machine Learning-Based Prediction. Journal of Scientific
Research and Reports, 29(4), 27–33.
Territorial and Temporal Variability: edit Fruits abiding by a https://doi.org/10.9734/jsrr/2023/v29i41741
tree limit cut more than once and, in many ways, clinch this
type of cut too by region (for example vary even through the 2. Bansal Y,Lillis D and Kechadi M.(2024)A neural meta
same regions of editing developed in Punjab and in Tamil model for predicting winter wheat crop yield.Machine
Nadu.). Geography and seasonal factors must be taken into Language.10.1007/s10994-023-06455-1113:6(3771-
consideration and this can be realized through the 3788).Online publication date: 1-Jun-2024.
development of specific models for the regions in concern.
3. Khade, s., Malaye, V., Mhase, S., Shitole, A., Shirsath, P.S.,
Adoption of a database which will enable changes of the
& Student (2023). Crop yield prediction using machine
models owing to users' inputs such as location state season
learning. International journal of advanced research in
will also improve prediction accurateness.
science, communication and technology.
Mathematical Models And Assessment: These machine
4. Kavita jhajharia, pratistha mathur, sanchit jain, sukriti
learning based models will involve training on the historical
nijhawan,crop yield prediction using machine learning and
surrender and rainfall data among other parameters of region
deep learning techniques,procedia computer science,volume
under cultivation, cropping patterns, pesticide use, and
218,2023,pages 406417,issn 1877-0509
fertilizer application. Therefore, one of the most important
issues will be successfully solving the task of model https://doi.Org/10.1016/j.Procs.2023.01.023
5. J. iajun Zong, Quanyin Zhu, “Apply Grey Prediction in the 16. Manideep, A. P. S., & Kharb, S. (2022). A comparative
Agriculture Production Price”, Fourth International analysis of machine learning prediction techniques for
Conference Communication and Informatics (ICCCI - crop yield prediction in India. Turkish Journal of
2014), Jan. 03 - 05, 2014, Coimbatore, INDIA. Computer and Mathematics Education
(TURCOMAT), 13(2), 120-133.
6. Dian T. Armanda, Jeroen B. Guinee and Arnold Tukker.
17. P. Sharma, P. Dadheech, N. Aneja and S. Aneja,
(2019). “The second green revolution: Innovative urban
"Predicting Agriculture Yields Based on Machine
agriculture's contribution to food security and sustainability
Learning Using Regression and Deep Learning," in
– A review” Global Food Security, Elsevier, 22 pp. 13-24,
IEEE Access, vol. 11, pp. 111255-111264, 2023, doi:
https://doi.org/10.1016/j.gfs.2019.08.002.
10.1109/ACCESS.2023.3321861.
7. Sajja, G. S., Jha, S. S., Mhamdi, H., Naved, M., Ray, S., & 18. Engen, M., Sandø, E., Sjølander, B. L. O., Arenberg, S.,
Phasinam, K. (2021). An investigation on crop yield Gupta, R., & Goodwin, M. (2021). Farm-scale crop
prediction using machine learning. In 2021 third yield prediction from multi-temporal data using deep
international conference on inventive research in computing hybrid neural networks. Agronomy, 11(12), 2576.
applications (ICIRCA). 19. Sekar, Shevanthe & E, Sathiyamoorthy. (2023). A
Https://doi.Org/10.1109/icirca51532.2021.9544815. SYSTEMATIC STUDY ABOUT THE CROP YIELD
PREDICTION WITH MACHINE LEARNING
8. 8 . Agarwal, s., & Tarar, S. (2021). A hybrid approach for
TECHNIQUES. Journal of Survey in Fisheries
crop yield prediction using machine learning and deep
Sciences. 2477-2496. 10.53555/sfs.v10i1.489.
learning algorithms. In journal of physics: conference series
(vol. 1714, no. 1, p. 012012). IOP publishing.
9. Khade, s., Malaye, V., Mhase, S., Shitole, A., Shirsath,
P.S., & Student (2023). Crop yield prediction using R.
Prabavathi, balika J chelliah , "A comprehensive review
on machine learning approaches for yield prediction
using essential soil nutrients," universal journal of
agricultural research, vol. 10, no. 3, pp. 288 - 303,
2022. Doi: 10.13189/ujar.2022.100310machine
learning. International journal of advanced research in
science, communication and technology.
10. Kavita jhajharia, pratistha mathur, sanchit jain, sukriti
nijhawan,crop yield prediction using machine learning
and deep learning techniques,procedia computer
science,volume 218,2023,pages 406417,issn 1877-
0509,https://doi.Org/10.1016/j.Procs.2023.01.023
11. Manivasagam, m. A., Sumalatha, P., Likitha, A.,
Pravallika, V., Satish, K. V., & Sreeram, S. (2022). An
efficient crop yield prediction using machine learning.
International journal of research in engineering, science
and management, 5(3), 106-111
12. Kalhotra, s. K. ., Prakash, k. ., Mishra , m. K. ., &
Kumar, m. S. A. K. . (2023). A study of crop yield
prediction using machine learning approaches. Journal
of advanced zoology, 44(S5), 1260–1263.
Https://doi.Org/10.17762/jaz.V44is-5.1192
13. Bansal y,lillis D and kechadi M.(2024)A neural meta
model for predicting winter wheat crop yield.Machine
language.10.1007/s10994-023-06455-1113:6(3771-
3788).Online publication date:
1-jun-2024.Https://dl.Acm.Org/doi/10.1007/s10994-
023-06455-1
14. Vikranth, K., & Prasad, K. K. (2024). Yield prediction
system–a systematic approach for predicting
agricultural commodity. ICTACT journal on soft
computing, 14(4).
15. Kamble, A., Rathod, K., & Hake, P. (2023). Survey on
Novel Approach for Crop Yield Prediction Using
Machine Learning. International Journal for
Research in Applied Science and Engineering
Technology.