Posted on 14 Nov 2023 — The copyright holder is the author/funder. All rights reserved. No reuse without permission. — https://doi.org/10.22541/au.169995548.

84464946/v1 — This is a preprint and has not been peer-reviewed. Data may be preliminary.

Smart Fields: Enhancing Agriculture with Machine Learning

Shivaraj S1 , Manju M1 , Rissi Kumar P1 , and Shivesh PR1
1
Rajalakshmi Engineering College

November 14, 2023

Abstract
Agriculture is a cornerstone of India’s economy, supporting a vast majority of its population. However, farmers grapple with
selecting the right crop due to diverse soil characteristics, environmental factors, plant diseases, and the need for consistent
crop monitoring. This paper presents a smart system assisting farmers in specific crop selection, integrating plant diseases
and consistent monitoring as vital features. By considering comprehensive data on environmental parameters(moisture), soil
characteristics (including N, P, K levels), plant diseases, and consistent crop monitoring, the system recommends the most
suitable crop for each season. Moreover, it offers fertilizer suggestions aligned with optimal nutrient requirements, particularly
focusing on N, P, and K levels, aiming to enhance farming efficiency and sustainability.

Manju M1, Rissi Kumar P2 , Shivesh PR3 , Shivaraj S4

[email protected] , [email protected] ,
[email protected] , [email protected]
Department of Artificial Intelligence and Machine Learning, Rajalakshmi Engineering college, Chennai,
India.
Abstract - Agriculture is a cornerstone of India’s economy, supporting a vast majority of its population.
However, farmers grapple with selecting the right crop due to diverse soil characteristics, environmental
factors, plant diseases, and the need for consistent crop monitoring. This paper presents a smart system as-
sisting farmers in specific crop selection, integrating plant diseases and consistent monitoring as vital features.
By considering comprehensive data on environmental parameters(moisture), soil characteristics (including
N, P, K levels), plant diseases, and consistent crop monitoring, the system recommends the most suitable
crop for each season. Moreover, it offers fertilizer suggestions aligned with optimal nutrient requirements,
particularly focusing on N, P, and K levels, aiming to enhance farming efficiency and sustainability.
Keywords – NPK levels, plant disease detection, Crop and fertilizer recommendation system.

1. INTRODUCTION
A considerable portion of India’s people receives their livelihoods from the agricultural sector, which also
significantly contributes to the country’s economic growth. However, farmers face a complex challenge in
determining the optimal crop for cultivation in a given context.
This complexity arises from the diverse soil characteristics, variable environmental conditions, prevalent
plant diseases, and the need for ongoing crop monitoring. To
navigate this intricate landscape, this paper presents a comprehensive and technologically advanced approach
to support decision-making in agriculture.

1
 Our proposed approach encompasses the use of a diverse range of advanced machine learning models, in-
cluding Random Forest model, Support Vector Regression (SVR) or Support Vector Classification (SVC)
based on the accuracy value , XGBoost, and Decision Tree
Posted on 14 Nov 2023 — The copyright holder is the author/funder. All rights reserved. No reuse without permission. — https://doi.org/10.22541/au.169995548.84464946/v1 — This is a preprint and has not been peer-reviewed. Data may be preliminary.

techniques. Through rigorous evaluation and comparison of these models, we aim to identify the optimal
model for precise crop detection. Our objective is to enhance accuracy and efficiency in determining the
most suitable crop for cultivation.
In the domain of plant disease detection, we harness the powerful capabilities of the ResNet model, a deep
learning architecture. Implementing this model enables us to accurately and swiftly detect and diagnose
plant diseases. This technology facilitates timely disease detection, empowering farmers to take proactive
measures to manage and mitigate disease outbreaks, thereby safeguarding their crops.
Furthermore, we envisage integrating IoT technology to conduct real-time soil tests and monitor critical soil
characteristics, including nutrient levels and moisture content. This technological advancement enhances our
understanding of soil health and facilitates
precise fertilizer recommendations tailored to the specific nutrient requirements of the soil. In the future, we
plan to expand this initiative to encompass real-time monitoring of soil attributes through IoT, presenting
an innovative and sustainable approach to farming practices.
This comprehensive approach aims to revolutionize agricultural practices by combining advanced machine
learning techniques, leveraging deep learning capabilities, and integrating IoT technology. Through in-
novative and technology-driven solutions, we hope to provide farmers with invaluable insights, promote
sustainable and efficient farming practices, and contribute to the advancement of India’s agricultural sector.

LITERATURE SURVEY
Machine learning is used in this integrated agricultural system to provide crop recommendations based
on historical planting data and soil nutrient analysis. It takes into account soil type and current weather
conditions to recommend optimal crops and nutrient adjustments. A smart irrigation system, which makes
use of IoT technology and machine learning, also monitors soil moisture levels. It assesses plant health in
real time using computer vision and determines whether irrigation is required. The system also checks the
likelihood of rain in the next 24 hours, saving water and avoiding over-irrigation. The Weather Underground
applet collects weather data and sends SMS alerts to users to help them make informed irrigation decisions,
ultimately promoting efficient crop cultivation and resource conservation.
By predicting crop suitability, this comprehensive agricultural system assists farmers in making informed
decisions. It improves accuracy by combining live data
with historical temperature, humidity, and rainfall data from government websites. The project combines
real-time field data from a DHT-22 sensor with soil type and historical weather data from government or
Google Weather API sources. On this dataset, supervised and unsupervised machine learning algorithms
are trained, and their accuracy is compared to deliver the most precise crop recommendations and fertilizer
suggestions to the end user via a responsive multilingual website. The DHT22 sensor, known for its precision
in monitoring temperature and humidity, and the Arduino Uno for data processing are among the hardware
components.
By combining real-time and historical data, this system assists farmers in making informed crop decisions. It
gathers data on temperature, humidity, and rainfall from government websites, as well as real-time field data
from a DHT22 sensor and user- entered soil type. The system forecasts weather conditions and crop suitabil-
ity using both supervised and unsupervised machine learning algorithms. The accuracy of Decision Tree, K-
NN, and Support Vector Machine models is compared in order to provide the best crop recommendation. A
responsive, multilingual website makes farmer interaction easier. Furthermore, the system provides fertilizer

2
 recommendations based on the crop selected. It collects and analyses data using IoT devices and cloud
platforms, ensuring precise crop cultivation guidance.
Posted on 14 Nov 2023 — The copyright holder is the author/funder. All rights reserved. No reuse without permission. — https://doi.org/10.22541/au.169995548.84464946/v1 — This is a preprint and has not been peer-reviewed. Data may be preliminary.

The system incorporates various sensors, such as the DHT11 for temperature and humidity, the MQ2 for
gas and smoke, the Soil Moisture Sensor, and the Light Intensity Sensor, all of which are placed in the field
to provide real-time data to a cloud server. This data is then processed and made available to users via
a website. Based on real-time sensor readings, the system determines the best crop for the field using the
K-Nearest Neighbors (KNN) machine
learning algorithm. Prediction is aided by a standardized dataset of crop requirements. For user convenience,
real-time data is plotted, and a Virtualization Page visualizes the data over time. Additionally, email
notifications are sent to users to keep them up to date on crop conditions.
[9] To improve farming procedures, the proposed smart agriculture system renders use of IoT and cloud
computing. It collects real-time field data employing sensors such as pH and moisture sensors, which is pro-
cessed by an ATMEGA328 microcontroller and transmitted to a cloud server via an ESP8266 Wi-Fi module.
Based on a pre-defined database, a user-friendly mobile app then provides farmers with valuable insights like
soil pH, moisture levels, and crop recommendations. This system enhances crop yield, decreases fertilizer
costs, and eliminates the need for manual soil testing. It promotes efficient farming practices, utilization of
resources, and agricultural sustainability by enabling data-driven decisions in order to meet the challenges
of a growing population and climate change. [16] The proposed method uses a sensor network to estimate
soil nutrients efficiently, assisting in the prediction of the best crop for the tested soil. Farmers connect their
NPK sensors to a centralized server. These sensors gather nutrient data from soil samples and send it to a
server via a Raspberry Pi. An algorithm (detailed in the following section) makes predictions based on sensor
data and historical information. When the analysis is finished, a message with the recommended crop is
sent to the registered farmer, allowing for more informed crop selection decisions. By providing personalized
crop recommendations based on real-time soil data, this system improves farming practices.

METHODOLOGY
Data Collection:
Initially, we collect a variety of environmental data such as temperature and rainfall, as well as soil parameters
such as nitrogen levels and pH. At the same time, we collect data on plant diseases and crop performance.
This diverse data is then combined into a uniform database, laying the groundwork for later deep research.

Data Preprocessing:
To prepare the collected crop images for training, essential preprocessing steps were carried out. This included
standardizing image dimensions, optimizing lighting and contrast, and augmenting the dataset through
rotations and scaling. These steps were crucial in enabling the model to effectively identify soil characteristics
and make accurate predictions. Additionally, the fertilizer dataset underwent various processing steps to
enhance accuracy in fertilizer detection.

Machine Learning Model Evaluation and Selection:

Examine and evaluate a variety of machine learning models like Random Forest, XGBoost, and Decision
Tree to efficiently detect and suggest crops. Use cross-validation and appropriate measures such as accuracy,
recall (sensitivity), specificity, F1 score and Confusion matrix.

3
 Deep Learning Model for Plant Disease Detection:
Gather a large dataset of plant images, including both healthy and ill plants. Using the photos, train a deep
Posted on 14 Nov 2023 — The copyright holder is the author/funder. All rights reserved. No reuse without permission. — https://doi.org/10.22541/au.169995548.84464946/v1 — This is a preprint and has not been peer-reviewed. Data may be preliminary.

learning network, especially ResNet, to accurately identify and diagnose plant illnesses. To improve both
performance and efficiency, fine-tune the model.

IoT Integration for Soil Monitoring

Implement an IoT soil monitoring system with pH, moisture, and temperature sensors. Integrate these into
a framework for collecting real-time data and securely transmitting it to a cloud platform. Integrate the
system with a farmer database. Continuously monitor crops for changes and immediately notify farmers of
important changes in crop attributes for timely intervention and optimal output.

Predictive Analytics:
Using integrated environmental, soil, and disease data, use machine learning models to identify acceptable
crops for the future crop cycle. Improve agricultural productivity and resource utilization by considering
optimal timings for irrigation, herbicide, fertilizer, and manure application.

Validation and Testing:

Validate the integrated system through rigorous real-world testing, including farmers and agricultural spe-
cialists actively to acquire diverse and insightful input. Conduct rigorous testing techniques to assess crop
recommendation accuracy and efficiency, disease detection, and overall operational efficacy. Use their com-
ments to make essential improvements, assuring the system’s dependability and usefulness in real-world
agricultural circumstances.

CONCLUSION
In conclusion, the development of an intelligent crop selection and agricultural decision support system is a
complex and multifaceted process. The objective is to provide farmers with accurate insights for informed
decision-making, from data collection and preprocessing through the integration of sophisticated technologies
such as IoT and machine learning. The
effectiveness of the system is proven by real- world testing, which includes feedback from farmers and agri-
cultural professionals. This iterative technique enables continual development, maintaining the system’s
accuracy, efficiency, and relevance in answering contemporary agriculture’s developing demands. Finally, the
incorporation of cutting- edge technology holds immense potential for revolutionizing agriculture, increasing
crop output, optimizing resource utilization, and encouraging sustainable agricultural practices.

References
F. K. Syed, A. Paul, A. Kumar, and J. Cherukuri, ”Low-cost IOT+ML design for smart farming with mul-
tiple applications,” in 2019 10th International Conference on Computing, Communication and Networking
Technologies, 2019.
A. Gupta, D. Nagda, P. Nikhare, and A. Sandbhor, ”Smart Management of Crop Cultivation using IoT and
Machine
Learning,” in 2019 IEEE.

4
 1. N. Radhika and Narendiran, ”Kind of Crops and Small Plants Prediction using IoT with Machine
Learning,” International Journal of Computer & Mathematical Sciences, 2018.
2. T Raghav Kumar, Bhagavatula Aiswarya, Aashish Suresh, Drishti Jain, Natesh Balaji, Varshini
Posted on 14 Nov 2023 — The copyright holder is the author/funder. All rights reserved. No reuse without permission. — https://doi.org/10.22541/au.169995548.84464946/v1 — This is a preprint and has not been peer-reviewed. Data may be preliminary.

Sankaran, “Smart Management of Crop Cultivation using IOT and Machine Learning,” International
Research Journal of Engineering and Technology (IRJET) Nov 2018, pp. 845850
3. R. Holambe, P. Patil, P. Pawar, S. Salunkhe, and H. Joshi, ”IOT based Crop Recommendation, Crop
Disease Prediction and Its Solution,” IRJET, 2019.
4. T. M. Mitchell, ”Machine Learning,” India Edition 2013, McGrawHill Education.
5. Kumar, T. R., Aiswarya, B., Suresh, A., Jain, D., Balaji, N., & Sankaran, V. (2018).
Smart management of crop cultivation using IOT and machine learning. International Research Journal of
Engineering and Technology (IRJET).
1. S. Mhaiskar, C. Patil, P. Wadhai, A. Patil, V. Deshmukh, ”A Survey onPredicting Suitable Crops for
Cultivation Using IoT,” International Journal of Innovative Research in Computer and Communication
Engineering.(2017)
2. Bondre, D. A., & Mahagaonkar, S. (2019). Prediction of crop yield and fertilizer recommendation using
machine learning algorithms.International Journal of Engineering Applied Sciences and Technology ,
4 (5), 371-376.
3. Bodake, Komal, Rutuja Ghate, Himanshi Doshi, Priyanka Jadhav, and Balasaheb Tarle. ”Soil based
fertilizer recommendation system using Internet of Things.” MVP Journal of Engineering Sciences 1,
no. 1 (2018): 13-19.
4. Khanal, Sami, Kushal Kc, John P. Fulton, Scott Shearer, and Erdal Ozkan. ”Remote sensing in
agriculture—accomplishments, limitations, and opportunities.” Remote Sensing 12, no. 22 (2020):
3783.
[12]. Velmurugan, S. ”An IOT based smart irrigation system using soil moisture and weather prediction.”
(2020).
1. Lokesh K, Shakti J, Sneha Wilson, Tharini MS, ”Automated Crop Prediction Based on Efficient Soil
Nutrient Estimation Using Sensor Network,” in National Conference on Product Design (NCPD 2016),
July 2016.
2. Ashokkumar, K., Chowdary, D. D., & Sree, C. D. (2019, October). Data analysis and prediction
on cloud computing for enhancing productivity in agriculture. In IOP Conference Series: Materials
Science and Engineering (Vol. 590, No. 1, p. 012014). IOP Publishing.
3. M. M. Ozguven and K. Adem, ”Automatic Detection and Classification of Leaf Spot Disease in Sugar
Beet Using Deep Learning Algorithms,” Phys. A, Stat. Mech. Appl., vol. 535, Dec. 2019, Art. no.
122537.
4. Shaikh, T. A., Rasool, T., & Lone, F. R. (2022). Towards leveraging the role of machine learning
and artificial intelligence in precision agriculture and smart farming. Computers and Electronics in
Agriculture , 198 , 107119.
5. Saranya, T., C. Deisy, S. Sridevi, and Kalaiarasi Sonai Muthu Anbananthen. ”A comparative study of
deep learning and Internet of Things for precision agriculture.” Engineering Applications of Artificial
Intelligence 122 (2023): 106034.
6. Roy, A. M., & Bhaduri, J. (2021). A deep learning enabled multi-class plant disease detection model
based on computer vision. Ai ,2 (3), 413-428.
7. Raviraja, S., K. V. Raghavender, Prashant Sunagar, R. K. Ragavapriya, M. Jogendra Kumar, and V.
G. Bharath. ”Machine learning based mobile applications for autonomous fertilizer suggestion.” In
2022 4th International Conference on Inventive Research in Computing Applications (ICIRCA) , pp.
868-874. IEEE, 2022.
8. Food and Agricultural Organization. (2019). The state of Food and Agriculture. 23- 27