Machine learning based farming

model for crop prediction using

weather forecast support

Guided by Prof. Atul Barve | Team Members: Aditi Pande, Aastha More, Anamika Rajput, Diya
Pancheshwar, Amit Sahu, Devansh Yadav | Department of Computer Science and Engineering, Prestige
Institute of Engineering Management and Research (PIEMR)

ABSTRACT
Agriculture, a cornerstone of human civilization, involves cultivating crops and raising livestock to
meet sustenance and economic needs. It plays a vital role in a nation's sustainability and has been a
catalyst for technological advancements. Modern methods like machine learning and deep learning
are revolutionizing agriculture by enabling higher crop yields with reduced water use and
environmental impact. A crop's growth is profoundly influe nced by climatic factors such as wind
speed, temperature, rainfall, and humidity, making weather prediction crucial for successful
farming. Machine learning, with its ability to process and learn from diverse datasets, has emerged
as a powerful tool to predict crop yields based on climate data. By leveraging such innovations,
farmers can make informed decisions, enhance productivity, and ensure sustainable agricultural
practices. Keywords: Crop yield prediction, Machine learning, Climate data, Sustainable farming.

Keywords-Crop yield prediction, Machine learning, Geographical Information, Agricultural

dynamic

I.INTRODUCTION: specifically neural networks, in mitigating these

challenges.
Weather fluctuations and their subsequent impact
In the realm of agriculture, accurate predictions
on crop yields are intricately linked phenomena
regarding weather patterns and crop yields hold
that significantly influence agricultural
paramount importance.
productivity and food availability. Accurate
The ability to forecast weather conditions and crop
weather predictions empower farmers to make
production outcomes plays a pivotal role in shaping
informed decisions regarding planting schedules,
agricultural practices, informing decision-making
irrigation practices, pest management strategies,
processes, and ultimately ensuring food security on
and harvesting timelines. By anticipating weather-
a global scale. In this section, we will delve into
related risks such as droughts, floods, heatwaves,
the significance of accurate weather and crop yield
and storms, farmers can adopt proactive measures
prediction, the challenges encountered by farmers
to safeguard their crops and minimize potential
and policymakers, and the role of artificial
losses. Additionally, precise forecasts enable
intelligence (AI) and machine learning (ML),
policymakers to formulate effective agricultural
policies, allocate resources efficiently, and mitigate operations. Analysis using these collected data
the adverse effects of climate variability on food makes data cleaning and preprocessing more
production. critical while using machine learning algorithms.
Moreover, reliable predictions of crop yields are Machine learning algorithms can analyze vast
indispensable for optimizing resource allocation, amounts of data from IoT sensors and other
enhancing supply chain management, and ensuring sources. It is a rapidly growing field that has the
market stability. By anticipating variations in crop potential to transform the way we predict and
output, stakeholders across the agricultural value analyze crop growth and output. Machine learning
chain can streamline distribution channels, manage algorithms use statistical/mathematical models and
inventory levels, and plan marketing strategies algorithms to analyze data and make predictions,
enabling computer systems to learn and improve
accordingly. Furthermore, accurate crop yield
from experience without being explicitly
forecasts facilitate risk assessment and financial
programmed. In agriculture, especially in the
planning for farmers, lending institutions, and cultivation area, machine learning algorithms can
insurance providers. Overall, the ability to forecast be trained on comprehensive data collected from
weather patterns and crop yields with precision is farms, such as weather patterns, soil properties,
instrumental in promoting agricultural crop growth stages, and pest and disease outbreaks.
sustainability, economic resilience, and food By evaluating the collected data, machine learning
security worldwide. models can forecast reap growth, output, and
In recent years, machine learning applications have quality with high accuracy.
entered our lives in many areas, from health to
defense industries and education to urbanization,
and have taken an effective way in decision-
making situations. At the same time, it started to
produce information and technology solutions by
forming the basis of the newly emerging search
engine infrastructure, such as ChatGPT (Chat
Generative Pretrained Transformer from Google
Bard, and similar AI-based chatbots and some
other tools). Many research companies reveal that
new trends will grow even more in various
platforms. In this respect, the effect of machine
learning-oriented systems and solutions in the
technology field will increase its effectiveness as a
huge multiplier, and many sectors, such as chip
design and traffic estimations, would be changed
by enforcing machine learning models.
Generally, it is essential to collect and analyze
accurate data using machine learning algorithms.
The data collection is critical in both quality and
size to obtain accurate results and make high
predictions. In general, big data has size, speed,
and various characteristics. Their large size helps FIGURE 1. Average yearly crop production 1997-2020 .

eliminate randomness and allows the data to

provide detailed results. In addition, large-scale
analysis data could be more structured. Using more The crop production for India from 1997 to 2020 is
than one dataset from different sources in the shown in Figure 1. From the analysis shown, it can
analysis will provide a higher success rate. Many be observed that the majority of the land in the
sources, such as sensors, social media, digital country was planted with wheat and rice, which
networks, physical devices, the stock market, and produced more than 73% of the nation's staple
health centers, are sufficient data sources. This data grains.
can be accessed through APIs, web collection, and India accounts for around 40 percent of global rice
direct access paths. Data can be in two forms: (Basmati and Non-Basmati) trade and exports to
static datasets or stream data. Data from different more than 150 countries. Trade ministry data
platforms are incorporated into the data processing showed that exports rose
Percent to 2.16 million tons in the first half of in making decisions related to irrigation, planting,
2022-23. and harvesting, thus optimizing yields.

Similarly, Ghimire et al. (2020) investigated the

II. Literature Review: influence of weather patterns such as seasonal
temperature and rainfall on crop performance,
The use of machine learning (ML) in agriculture finding that a combination of historical crop yield
has gained significant attention over the last decade data and weather forecasts improved predictive
due to its ability to predict crop yields, optimize models. The study concluded that weather data not
agricultural practices, and mitigate risks arising only improves crop yield predictions but also helps
from climatic variations. Integrating weather in predicting the potential for crop diseases that
forecasts into these ML models enhances arise due to specific climatic conditions.
prediction accuracy, as crop performance is heavily
influenced by environmental conditions like 3. Machine Learning Models Using
temperature, precipitation, and humidity. Weather Forecasts
1. Machine Learning in Agriculture Incorporating weather forecasts into machine
learning models has proven to significantly
Machine learning has been applied to a wide range improve crop prediction accuracy. Cheng et al.
of agricultural tasks, particularly in the prediction (2022) developed a decision support system (DSS)
of crop yields. One of the key contributions is in that combined weather forecasts with crop-specific
the prediction of crop yield using data such as soil growth models to offer real-time advice to farmers.
properties, historical yields, and climate data. The model used weather forecasts to suggest
Khosla et al. (2019) demonstrated that machine optimal planting times and identified potential risks
learning models, including random forests and such as drought or frost, allowing farmers to adapt
support vector machines (SVM), can effectively their strategies accordingly.
predict crop yields by integrating weather and soil
parameters. The study showed that data from Machine learning models that integrate weather
sensors, alongside weather forecasts, led to data, such as ensemble models, have been used to
improved accuracy in yield prediction. predict crop growth under varying environmental
conditions. Narayan et al. (2019) used weather
Additionally, Zhang et al. (2021) used machine parameters and soil data in a support vector
learning to predict crop growth stages, improving machine (SVM) model for predicting the yield of
the efficiency of farm operations. By integrating maize and wheat. This hybrid model showed
data from satellite imagery and weather forecasts, improved performance compared to traditional
the study improved the prediction of growth stages models, demonstrating the added value of weather
such as flowering and maturation, which are forecasts in enhancing yield prediction accuracy.
essential for timely harvest planning.
4. Challenges in Integrating Machine Learning and
2. Weather Data in Crop Prediction Weather Forecasts
Despite the promise of combining weather
Models forecasts with machine learning models for crop
prediction, several challenges exist. One significant
Weather data plays a crucial role in determining issue is the availability and quality of data. Raza et
the success of agricultural activities. Several al. (2021) noted that weather data can be
studies have focused on incorporating real-time inconsistent, particularly in remote agricultural
weather forecasts into crop prediction models to areas, where data from meteorological stations may
enhance their precision. Basso et al. (2018) be sparse or unreliable. This can affect the
explored the integration of short-term weather accuracy of the predictions generated by machine
forecasts (temperature, rainfall, etc.) into crop learning models.
growth models. The authors emphasized that
accurate weather predictions could guide farmers Another challenge is the need for real-time data
updates. Crop growth is a dynamic process
influenced by rapidly changing weather conditions.
Bennett et al. (2021) highlighted the difficulty in
maintaining accurate predictions with static
models, emphasizing the need for real-time
weather data integration to update forecasts
continuously.

Moreover, the generalization of models across

different geographical regions remains a challenge.
Mao et al. (2020) pointed out that machine learning
models trained on data from one region often fail
to perform well in different climatic zones due to
differences in weather patterns and soil conditions.
They suggested that more flexible and adaptive
ML models are required to handle this variability.

5. Future Directions
The integration of deep learning models into
agricultural predictions is a promising future
direction. Zhang et al. (2020) explored the use of
convolutional neural networks (CNNs) to process
satellite images in conjunction with weather data
for more accurate crop health monitoring. Such
models allow for the detection of diseases, pests,
IV. Challenges Faced by
and growth stages, facilitating timely interventions Farmers and Policymakers:
and better crop management.
Despite the critical importance of weather and crop
Furthermore, the advent of IoT-based sensors and yield prediction, farmers and policymakers
edge computing can improve the accessibility of encounter numerous challenges in accurately
real-time weather and environmental data. forecasting these variables. One of the primary
challenges stems from the inherent unpredictability
of weather patterns, which are influenced by a
III. ER Diagram: myriad of factors such as atmospheric conditions,
ocean currents, and geographic features. Variability
in climate phenomena, including El Niño, La Niña,
and the North Atlantic Oscillation, further
complicates forecasting efforts, leading to
uncertainties in precipitation, temperature, and
humidity levels.
The dynamic nature of agricultural systems
introduces complexities that pose challenges for
crop yield prediction. infestations, and crop
diseases interact in intricate ways, making it
challenging to model and predict crop production
outcomes accurately.
Moreover, socioeconomic factors such as market
demand, trade policies, and labor availability add
another layer of complexity to the forecasting
process, influencing planting decisions, yield
projections, and market prices. Furthermore,
traditional forecasting methods often rely on
simplistic statistical models that fail to capture the
nonlinear relationships and complex interactions up-to-date forecasts that are essential for timely
inherent in weather and crop systems. decision-making in agriculture. Artificial
These models may overlook critical variables, intelligence and machine learning, particularly
underestimate uncertainty, and yield inaccurate
neural networks, hold immense promise for
predictions, thereby limiting their practical utility
revolutionizing weather and crop yield prediction
for farmers and policymakers alike. As a result,
there is a pressing need for advanced predictive in agriculture. By addressing the challenges of
analytics techniques that can effectively harness uncertainty, complexity, and variability inherent in
the wealth of available data and provide reliable weather and crop systems, these advanced
forecasts for weather and crop yields. techniques can empower farmers, policymakers,
and other stakeholders with actionable insights that
contribute to agricultural sustainability, economic
V. Role of Artificial
stability, and food security. In the following
Intelligence and Machine sections, We will delve deeper into the
Learning: applications, methodologies, and case studies that
illustrate the transformative potential of AI and
In recent years, artificial intelligence and machine ML in weather and crop yield prediction.
learning have emerged as powerful tools for
addressing the challenges associated with weather
Random Forest (RF):
and crop yield prediction. Unlike traditional
statistical methods, which rely on predefined
The RF technique is a perfect example of ensemble
assumptions and linear models, AI and ML learning in action since it connects several
techniques, particularly neural networks, offer a classifiers to tackle the challenging problem and
data-driven approach that can uncover hidden improve a model’s efficiency. The “forest” created
patterns, identify complex relationships, and adapt with this approach is actually a collection of
to changing conditions. Neural networks,inspired decision trees. In each decision split, RF
characteristics are chosen at random. Picking traits
by the structure and function of the human brain,
that encourage prediction and lead to increased
excel at processing vast amounts of efficiency reduces the correlation across trees. The
data, learning from examples, and making Random Forest ML classification approach
predictions with remarkable accuracy. By generates the final output by combining the results
leveraging large datasets containing meteorological of all the decision trees after segmenting the
variables, soil properties, historical yield data, and dataset into smaller subsets or trees. The Bagging
subcategory of ensemble learning methods includes
agronomic practices, neural networks can generate
Random Forest. A Sample of rows and features
forecasts for weather patterns and crop yields with from the primary dataset are selected at random
unprecedented precision. The ability of neural and fed into the Random Forest Technique’s
networks to capture nonlinear decision trees. It can also carry out jobs requiring
interactions,generalize patterns across different both regression and classification. It also works
geographical regions, and adapt to evolving well with huge, highly dimensional data sets, and
most significantly, it greatly improves the model’s
environmental conditions makes them well-suited
accuracy and fixes the overfitting problem.
for agricultural prediction tasks. Moreover, the
integration of AI and ML techniques into weather
and crop yield prediction models enables
continuous learning and refinement, allowing these
models to improve over time and adapt to changing
scenarios. By incorporating real-time observations
from weather stations, satellites, and remote
sensing technologies, neural networks can provide
 process of choosing a subset of the relevant
features from available data. For Crop Prediction,
following features were selected: Nitrogen,
Phosphorous, Potassium, Temperature, Humidity,
pH and Rainfall. For Yield Prediction, features
selected are as follows: Region, soil type,
temperature in celsius, weather condition and the
crop.

VIII. Training and

evaluation of models:

The crop prediction uses the multi-class

classification machine learning model to predict
the crop for a set of given input features. Whereas
the yield prediction incorporates the regression
VI. Datasets: model to predict the yield for a given set of input
features. For training and evaluation of models,
The public datasets have been chosen because they Google Colab Platform is used. While the User
are readily available and easily accessible. Kaggle Interface for the project is built using ReactJs, the
is a popular platform for finding and sharing backend is built using Python Flask framework.
datasets, so we were able to find datasets that met
our criteria. We selected the following datasets
namely:
IX. Application and
Advantages over existing
versions:
Agriculture Crop Yield:
The model can be used to create an impact on right
This dataset contains agricultural data for crop selection as the user would get fair predictions
1,000,000 samples aimed at predicting crop yield on yield as well as crop. Also yield prediction
(in tons per hectare) based on various factors. The would be important in financial assessment of crop
dataset can be used for regression tasks in machine strategy. Model is useful if the user wants to
learning, especially for predicting crop compare yield for multiple crop options and then
productivity. select the best one. It could also be used in a wide
geography to estimate the yield for a particular
crop. This project can be used directly by end users
as farmers for taking predictions for their
conditions. Instead, it can also be used by
government agencies for planning and policy
making if modified with wider access to reliable
closed source government data. It can also be used
by NGOs which work for educating farmers in
adopting new technologies and precision
agriculture. Also, it can be used in fields where
monetary calculations come into picture as it is
dependent on how much yield could be produced
like in insurance claims or loan policies.
VII. Feature Selection:
The project improves the prediction accuracy by
suitable data gathering, cleaning and selecting the
A machine learning model’s performance can be best accurate model. Also, the project incorporates
improved through feature selection, which is the both crop as well as yield prediction. So, the
project is using classification as well as regression algorithm yields 88.50%, and the rest of the
models for necessary functionality. It adds value to classification algorithms have more than 90%
the modern agriculture setup by providing a way to accuracy. Error metrics can be formalized as
add to the reliability of crop selection which in turn follows:
improves the yield and financial stability.
K (Kappa) = (Po − Pe)/(1 − Pe)
where Po is relative observed agreement and Pe is
hypothetical probability of chance agreement.
X. Experimental Results:
MAE(Mean Absolute )𝑣𝑎𝑙𝑢𝑒𝐼=∑Ni=1|
In this work, we used 15 different machine learning
yi−xi|n
algorithms to model agriculture data, which where n is the total number of data, xi is the true
recommended to farmers the most suitable crops to value, and yi is the prediction.
produce on the farm. Tabular data are used in this
work for classification of the crop data. Data is
collected from the Kaggle database, which is an
RMSE (Root Mean Square
online platform for scientists to share their research Error)=[∑ni=1(xi−x′i)]/n where xi is
data. The dataset includes several features, such as observed and xi′ is the predictive value.
ratio of Nitrogen content (N), temperature, pH
value of the soil, rainfall, humidity, ratio of
RAE
Phosphorous content (K), and ratio of Potassium
content (P) in the soil. The crop prediction dataset (Relative Absolute Error)=∑ni=1|
has 2200 records, which have 22 crop labels The yi−y′i|∑ni=1|yi−y|
dataset includes several features, such as the ratio
where y is the average value of the data.
of Nitrogen content (N), temperature, pH value of
the soil, rainfall, humidity, ratio of Phosphorous
content (K), and the ratio of Potassium content (P) Root Relative Squared Error
in the soil. Sixty-seven percent of the data is used
in training, and the rest of the data are used for
(RRSE)=A=∑ni=1(Pi−Tj)2∑ni=1(Ti−T)
testing. 2⎷
where P is the predicted value and T is the target
value

XI. Web application:

A simple web application is also created that

outputs the prediction. This application assists
farmers in choosing which crop to cultivate to
elicit the greatest return. For this purpose,
Tandojam’s profusely producing crops are se-
lected and their names are foreseen. Utilizing
XGBoost (i.e., achieved the highest prediction
accuracy), the prepro-cessed dataset is
trained.Prediction is done using the immediate
weather data for the chosen district that is acces-
sible via API. For the chosen district, the trained
model produced the appropriate crop prediction.
It shows accuracy values and error rates for the
algorithms considered. Bayes Net, Naïve Bayes
Classifier, Hoeffding Tree, and Random Forest
algorithms yield the best accuracy. The DT
 XII. Conclusion:

Our research highlighted the significance of

incorporating machine learning algorithms in
modern agriculture to optimize reap production.
This study identifies the challenges and
opportunities associated with integrating these
technologies in agriculture. It presents
experimental results that demonstrate the impact of
changing labels on the accuracy of data analysis
algorithms along with accuracy, error values, build,
and test time for each classification algorithm. The
findings suggest that analyzing wide-ranging data
collected from farms, including real-time data from
datasets, can enable farmers to make more
informed decisions about factors that affect harvest
growth. Despite the challenges associated with
deploying machine learning in agriculture, our
results achieved so far are very promising in that
machine learning approaches will become
increasingly crucial for production predictions in
agriculture in the future. In this experiment, crops
were investigated according to general
characteristics using different machine learning
algorithms, and valuable results were obtained by
making predictions in cases where certain crop
types are unknown or cannot be easily identified.
Our work indicated that appropriate feature
selection is critical to achieve better accuracy in
machine learning algorithms while analyzing
agricultural data. Using the variety of features in
the dataset, it achieved the highest accuracy,
reaching 93.2% with Random Forest(). This
research provided valuable insights into the
potential benefits of these technologies in modern
agriculture, and further research and development
in this field could help optimize crop production,
reduce waste, and improve food security globally.
The city of the user taken as an input is given to the
API call as a parameter. The temperature and
humidity fields from API response are given to the
crop prediction model as input along with other
data. This helps the system to give real-time
predictions. “OpenWeatherMap” API is used for learning models with multi-source
the same. For creating an API URL, base URL and environmental data to predict wheat yield
API key is used which is unique with each in China. Comput. Electron. Agric. 2022,
subscription. User’s city name is passed in the 194, 106790. [Google Scholar] [CrossRef]
complete URL as a parameter and response is 2. Kuradusenge, M.; Hitimana, E.;
collected. From the collected response, required Hanyurwimfura, D.; Rukundo, P.;
fields i.e., temperature and humidity are passed to Mtonga, K.; Mukasine, A.; Uwitonze, C.;
the ML model for predicting crop. Crop yield Ngabonziza, J.; Uwamahoro, A. Crop
prediction is a complex process which relies on Yield Prediction Using Machine Learning
several different factors including weather, soil, Models: Case of Irish Potato and Maize.
fertilizers, pest infestations, etc. In this paper, we Agriculture 2023, 13, 225. [Google
predict the crop yield using weather and soil Scholar] [CrossRef]
parameters. Calculating feature importance for a 3. Gehlot, A.; Sidana, N.; Jawale, D.; Jain,
Random Forest Regressor with one-hot encoded N.; Singh, B.P.; Singh, B. Technical
features involves determining the contribution of analysis of crop production prediction
each feature to the model’s predictive performance. using Machine Learning and Deep
It is done through the following steps. Train the Learning Algorithms. In Proceedings of
Random Forest Regressor, Access Feature the International Conference on
Importances: Random Forest Regressor has an Innovative Computing, Intelligent
attribute named feature_importances_. Map Communication and Smart Electrical
Feature Importances to Original Features: Every Systems (ICSES), Chennai, India, 24–25
one-hot encoded feature is mapped with its original September 2022; pp. 1–5. [Google
feature. Aggregate Feature Importances: By Scholar]
aggregating we get every categorical feature’s 4. Cui, Y.W.; Henrickson, K.; Ke, R.; Pu, Z.;
importance, Rank Feature Importances in Wang, Y. Traffic graph convolutional
descending order of importance. recurrent neural network: A deep learning
First, datasets are loaded and cleaned from framework for network-scale traffic
insignificant features. After Data Preparation, data learning and forecasting. IEEE Trans.
is split into training and testing data and various Intell. Transp. Syst. 2019, 21, 4883–4894.
models are fitted and tested for accuracy. Feature [Google Scholar] [CrossRef]
Importance is calculated to determine the relative
5. Rashid, M.; Bari, B.S.; Yusup, Y.;
significance or contribution of individual features
Kamaruddin, M.A.; Khan, N. A
in the ML model. For the crop prediction model,
Comprehensive Review of Crop Yield
Drop Column Importance, also called as
Prediction Using Machine Learning
“permutation importance” or “feature importance
Approaches with Special Emphasis on
by feature shuffling,” is calculated. The project
Palm Oil Yield Prediction. IEEE Access 2021, 9,
improves the prediction accuracy by suitable data 63406–63439. [Google Scholar]
gathering, cleaning and selecting the best accurate [CrossRef]
model. 6. Venugopal, Anakha, S. Aparna, Jinsu
Mani, Rima Mathew and Vinu Williams.
”Crop yield prediction using machine
learning algorithms.” International journal
XIII. References: of engineering research & technology
(IJERT) NCREIS 9, no. 13 (2021)
1. Li, L.; Wang, B.; Feng, P.; Liu, D.L.; He, 7. Taheri, M.; Schreiner, H.K.;
Q.; Zhang, Y.; Wang, Y.; Li, S.; Lu, X.; Mohammadian, A.; Shirkhani, H.; Payeur,
Yue, C.; et al. Developing machine P.; Imanian, H.; Cobo, J.H. A Review of
Machine Learning Approaches to Soil
 Temperature Estimation. Sustainability
2023, 15, 7677. [Google Scholar]
[CrossRef]
8. D. K, R. M, S. V, P. N, and I. A.
Jayaraj(2021)," Meta-Learning Based
Adaptive Crop Price Prediction for
Agriculture Application," in 2021 IEEE,
5th International
Conference on Electronics,
Communication, and Aerospace
Technology (ICECA), pp. 396-402, DOI:
10.1109/ICECA52323.2021.9675891.
9. Ms. Fathima, Ms. Sowmya K, Ms. Sunita
Barker, Dr. Sanjeev Kulkarni(2020),
"Analysis of Crop yield Prediction using
Data Mining Technique" International
Research Journal of Engineering and
Technology (IRJET), Volume: 07 Issue:
05.
10. L’heureux, A.; Grolinger, K.; Elyamany,
H.; Capretz, M. Machine learning with
Big Data: Challenges and approaches.
IEEE Access 2017, 5, 7776–7797.
[Google Scholar]
[CrossRef]
11. Medar, Ramesh, S, Vijay, Shweta. “Crop
yield prediction using machine learning ”
International Journal of Advanced
Research in Computer Science and
Software Engineering, vol. 9, no. 5, pp. 1-
6, May 2019.CrossRef
12. Meshram V, Patil K, Meshram V,
Hanchate D, Ramkteke SD. Machine
learning in agriculture domain: A state-of-
art survey. Artificial Intelligence in the
Life Sciences. 2021 Dec
1;1:100010.CrossRef