Department of Computer Engineering

Academic Year 2023-24

LY Internal Assessment A
Krishi Mitra - Intelligent Crop Recommender System
Tirth Panchal A58

Supervisor: Dr. Sarita Ambadekar

Introduction
Indian farmers, through their relentless efforts and selflessness, have been the backbone of
the nation's economy, providing livelihoods to millions and feeding a vast population.

Today, we're here to explore a cutting-edge solution that's redefining agriculture: Crop
Recommendation Systems. These systems leverage data and technology to empower
farmers with informed decisions, maximizing yields, minimizing waste, and ensuring
sustainability. The challenges agriculture faces today:

Growing Global Population: With the world's population nearing 10 billion by 2050,
agriculture must find innovative ways to meet increasing food demand.
Climate Change: Erratic weather patterns and extreme temperatures threaten crop
production and food security.
Resource Scarcity: Depleting resources and limited access to quality inputs put
additional pressure on agriculture.

In response, Crop Recommendation Systems provide tailored guidance for crop selection,
cultivation techniques, and resource management. They incorporate factors like soil
characteristics, climate conditions, and historical data. These systems empower farmers,
enhance sustainability, and contribute to global food security.
Objectives :
In response to the pressing challenges faced by agriculture, including declining crop yields,
rampant plant diseases, and uncertain crop planning, we've devised a holistic solution. At its
core, this solution harnesses the power of AI-ML decision systems to provide a range of
critical features:

❖ Crop recommendation:Leveraging AI-ML algorithms to match soil and climate data

with suitable crop varieties using Random Forest.

❖ Economic Efficiency: Measure changes in crop yields and agricultural productivity

resulting from the crop recommendation system. Improved yields can lead to
higher economic efficiency.

❖ Environmental Responsibility: Minimize the environmental impact of excessive

fertilizer use by preventing nutrient runoff and contamination of water sources,
safeguarding local ecosystems.

❖ Improved Yield and Quality: Provide guidance on best practices for soil preparation,
irrigation, fertilization, and pest control to enhance crop yield, quality, and overall
production efficiency.
 Machine Learning-Based Crop Recommendations
for Precision Farming to Maximize Crop Yields
Authors: C. Sagana; M. Sangeetha; S. Savitha; K. Devendran; T.
Kavin; K. Kavinsri; P. Mithun

● Proposes Crop recommendation system.

● Comparison is done between various classifiers
including Decision Trees, Support Vector
Machines, Naive Bayes, Logistic Regression,
Random Forests, and XGBoost.
● Dataset is gathered from Kaggle, which
includes parameters such as humidity, rainfall,
soil PH, nitrogen (N), phosphorus (P), and
potassium (K).
● The data set contains 2200 instances of data.

Computer Engineering Department 4

Machine Learning-Based Crop Recommendations
for Precision Farming to Maximize Crop Yields
Authors: C. Sagana; M. Sangeetha; S. Savitha; K. Devendran; T.
Kavin; K. Kavinsri; P. Mithun

Computer Engineering Department 5

 Machine Learning-Based Crop Recommendations
for Precision Farming to Maximize Crop Yields
Authors: C. Sagana; M. Sangeetha; S. Savitha; K. Devendran; T.
Kavin; K. Kavinsri; P. Mithun
Conclusion
This paper proposes crop recommendation. Different machine learning algorithms were
compares based on accuracy metrics. The Random Forest algorithm outperforms all the other
algorithms based on accuracy of 96.34%.

Computer Engineering Department 6

 Crop Recommendation using Machine Learning
Techniques
Authors: Saifulla Shariff; M. Sobhana; A. H. L. Swaroop; V Phani
Kumar
● Web-based system for crop recommendation,
fertilizer advice, and plant disease prediction.
● The application aims to increase agricultural
productivity by addressing critical crop-related
issues.
● The models used in the project are trained and
tested on various datasets.
● Crop Recommendation: Based on the N P K,
temperature, humidity, and ph, the model will
recommend the optimum crop to grow on the
given soil.
● It also presents the different treatment options

Computer Engineering Department 7

 Crop Recommendation using Machine Learning
Techniques
Authors: Saifulla Shariff; M. Sobhana; A. H. L. Swaroop; V Phani
Kumar
Conclusion
This paper proposes crop recommendation, fertilizer recommendation and plant disease
identification based on machine learning algorithms.Gradient Boosting was the
most accurate of the algorithms we tested.

Computer Engineering Department 8

 Crop Recommendation System Using Ensembling Techniques
Autors: K Gnana Sandhya; Sandeep Vemuri; K Sai Deeksha; T Anvitha

● User-friendly mobile application for a crop

recommendation system.
● The app employs an Ensemble model that
combines KNN, Random Forest, Naïve Bayes,
Logistic Regression, and SVM.
● These models' decisions are combined using
Majority Voting.
● The dataset contains a CSV file with 2200
instances and 8 attributes related to soil and
weather.
● Through data preprocessing, the dataset is
reduced to 1728 instances.

Computer Engineering Department 9

 Crop Recommendation System Using Ensembling Techniques
Autors: K Gnana Sandhya; Sandeep Vemuri; K Sai Deeksha; T Anvitha

Majority Voting in Machine Learning:

● Ensemble Technique: It's a method used in ensemble learning to combine predictions from
multiple classification models.
● Classification Task: Typically used for classification tasks to predict the class or category of
input data.
● Multiple Models: Train several models using different algorithms or data subsets.
● Individual Predictions: Each model provides a class prediction for the input data.
● Voting: Count the votes for each class based on individual model predictions.
● Final Prediction: The class with the most votes becomes the final prediction. Ties can be
resolved in various ways.
● Improves Accuracy: Helps improve overall prediction accuracy and model robustness.
● Common in Voting Classifiers: Often used in "Voting Classifiers" with options like "Hard
Voting" (majority voting) and "Soft Voting" (averaging class probabilities).
● Combines Model Strengths: Useful when different models have different strengths and
weaknesses.
Computer Engineering Department 10
 Crop Recommendation System Using Ensembling Techniques
Authors: K Gnana Sandhya; Sandeep Vemuri; K Sai Deeksha; T Anvitha

Conclusion:
This paper proposes crop recommendation system based on ensemble framework
of various classifiers. Majority voting technique is used to generate output.These
ensemble framework, achieves an impressive accuracy of 99.6% using the
majority voting technique.

Computer Engineering Department 11

Literature Survey:
Sr. No. Title Author Methodology Advantage Limitations

1 Crop Recommendation Gaurav Chauhan, Alka Random forest and decision trees Timely notifications about crop Implementing and
System Using Machine Chaudhary to predict the crop and forecasts reach farmers. maintaining crop
Learning Algorithms performance compared using recommendation systems can
WEKA Tool be expensive, especially in
resource-constrained regions.

2 A Crop Recommendation S. Pudumalar, E. Ensemble model with majority By diversifying crops or Gathering and sharing
System for Precision Ramanuja voting technique using Random choosing those suitable for agricultural data may raise
Agriculture trees, CHAID,KNN and naive specific conditions, farmers can privacy concerns for farmers,
bayes reduce risks associated with which can limit data
weather fluctuations, pests, and collection and analysis.
diseases.

3 Machine Learning-Based C. Sagana, M. Naive Bayes, SVM, Logistic This will help to maximize the Can integrate plant disease
Crop Recommendations Sangeetha, S. Savitha Regression, Random Forest, yields of crops which increase prediction module as well
for Precision Farming to Decision Trees and XGBoost, for the profits of farmers. Based on
Maximize Crop Yields crop recommendations. Random comparison from different
Forest stands out as the most models RF gave more accuracy
accurate, achieving a 96.34% of 96.34%.
accuracy rate.
Literature Survey:
Sr. No. Paper Title Algorithm Used Advantages Limitation Limitations

7 Crop Recommendation with J Sreemathy, N Prasath BiLSTM-MERNN model effectively learn and no such specific interface
BiLSTM-MANN Algorithm for outperforms Multilayer understand complex or GUI which can be
Precision Agriculture Perceptron and CNN model for temporal patterns related used by farmers, can
crop recommendation system to weather conditions and integrate plant disease
soil properties, high identification and
accuracy, scalable to fertilizer
handle large dataset recommendation as well
8 An Enhanced Light GBM Model Namrata Bhatt, Sunita Light GBM model provides High efficiency, faster no such specific interface
with Data Analytical Approach for Varma more accuracy than logistic training, easily handle or GUI which can be
Crop Recommendation regression model for crop large dataset, high used by farmers, can
recommendation system. accuracy, high scalability integrate plant disease
identification and
fertilizer
recommendation as well

9 Crop Yield Prediction and Fertilizer K P K Devan; B Swetha; P Random Forest and Logistic High accuracy due to Hybrid models introduce
Recommendation System Using Uma Sruthi; S Varshini Regression hybrid model hybrid model complexity, have
Hybrid Machine Learning outperforms individual models overfitting risks and hard
Algorithms like RF,DCT,SVM,LR,etc. to maintain
Literature Survey:
Sr. No. Paper Title Algorithm Used Advantages Limitation Limitations

4 Crop Recommender System SHILPA MANGESH PANDE; Employs various algorithms like Recommending time of Lack of interpretability,
Using Machine Learning PREM KUMAR RAMESH; SVM,MLR,RF,KNN,ANN to fertilizer is a unique and potential overfitting,
Approach ANMOL ANMOL; B. R predict crops.RF outperforms useful module longer prediction times,
AISHWARYA; them all.Provides fertilizer high memory usage.
recommendation and
recommends at what time is
fertilizer to be used
5 AGRI-PRO: Crop, Fertilizer Yalabaka Srikanth, Meghana Compares various algorithms Provides services such as No multilingual support,
and Market Place Daddanala, Manchala Sushrith, such as DT,NB,SVM,KNN,RF marketplace and farm can integrate pant disease
Recommender for Farmers Pranith Akula, Ch. Rajendra and ANN for crop and fertilizer discovery . The random identification as well
using Machine Learning Prasad, Dasari Sindhu Sri recommendation modules. RF forest algorithm provides
Algorithms outperforms them all for both the higher accuracy than
the modules other models

6 KRISHI RAKSHAN - A D. N. V. S. L. S. Indira, M. RF is used for fertilizer predicts crop disease as No multilingual support.
Machine Learning based New Sobhana, A. H. L. Swaroop, V recommendation, XGB is used well, RF, XGB and
Recommendation System to the Phani Kumar for crop recommendation and mobile net outperforms
Farmer mobile net is used for crop existing classifiers
disease detection.
Literature Survey:
Sr. No. Paper Title Algorithm Used Advantages Limitation Limitations

10 An Efficient Machine Nidhi H Kulkarni; G N Real time data gathered using KNN is non-parametric and Can not handle large
Learning Approaches for Srinivasan; B M Sagar; N K Arduino uno and sensors , KNN thus simple to construct dataset
Crop Recommendation Cauvery Classifier used for crop
based on Soil prediction
Characteristics

11 Crop Recommendation and Anuraj Singh, Amit Kumar ANN is used for crop ANN can identify complex no such specific interface
Disease Detection using Bhamboo recommendation systems and 2d patterns and relationships or GUI which can be
Deep Neural Networks CNN is used for plant disease between agricultural data and used by farmers.can
detection systems. have high scalability. CNN are integrate fertilizer
specializer for images, recommendation as well
provides effective feature
extraction, high accuracy,
highly scalable and accurate
12 ENSEMBLED CROPIFY – K. Devi Priya; A. Sasi RF gives best accuracy for crop ResNet mitigate the Computationally
Crop & Fertilizer a System Samyogitha; A. Vamsi prediction, RF and SVM overfitting problem making it Expensive, less
with Leaf Disease Prediction Krishna Reddy; B. Divya Sri combined are use for Fertilizer easier to train Neural interpretability
recommendation , CNN with Networks and aids in
ResNet is used for leaf disease preventing vanishing gradient
prediction problem
Literature Survey:
Sr. Paper Title Algorithm Used Advantages Limitation Limitations
No.

13 Crop Recommendation using Shivesh Tiwari; Somesh RF emerged as most accurate Improved efficiency, better Complexity, increased
Machine Learning and Plant Kumar; Sunil Tyagi; algorithm for crop prediction and performance, parameter computational
Disease Identification using Minakshi Poonia For Plant Disease Identification efficiency, faster training, state- requirements due to
CNN and Transfer-Learning EfficientNetV2 outperformed of-the-art accuracy due to EfficientNetV2
Approach other CNN architectures EfficientNetV2

14 A Machine Learning-based Jeeva Ganesh R; Harish Government dataset is used, Ensemble models are use Increased training time , is
Approach for Crop Yield D; Priya B AdaBoot used for crop which grant stability , reduce more complex and
Prediction and Fertilizer recommendation and Fr used for overfitting, make better computationally expensive
Recommendation fertilizer recommendation predictions thus increasing
accuracy

15 Using A Two-Fold Machine Ginka Hemdeep; Crop prediction and fertilizer Ensemble models are used Can integrate plant disease
Learning Approach for Crop Anumukonda Sowjanya; . recommendation is done using which help increase accuracy prediction module as well
And Fertilizer Matilda Florence DCT,NB,LR,RF,SVM,XGB. and give better results overall
Recommendation System Comparative analysis was
performed

16 Machine Learning based Crop Mohammed Abdul Model deployed on cloud for A cloud platform allows Cloud Disadvantages
Recommendation on a Cloud Raheem; Mohammed faster processing,RF gives best organizations to rapidly deploy include Downtime, data
Sabir Hussain; Syed accuracy out of DCT,LR,XGB applications and services, and security, cost, vendor lock-
Ayaan Ahmed provides the ability to store and in, latency, limited control.
analyze large amounts of data.
Future Scope:
● In future work, project can be extended to classify plant diseases using CNN and provide remedies for them.

● The Web app can be made in local languages to make it more accessible to people

● Suggestions based on amount of irrigation can also be included.

● Inclusion of pest detection and suggesting treatments to minimize the damage in crop yield can also be carried
out.

● Efficient load balancing algorithms can be implemented for better traffic distribution in the deployment.

● Considering more external parameters with soil data to improve results and exploring alternative methodologies
such as deep learning techniques and hybrid models
System Architecture:
Hardware Requirements:
• Processor: 2 gigahertz (GHz) or faster processor or SoC.

• RAM: 6 gigabyte (GB) for 32-bit or 8 GB for 64-bit.

• Hard disk space: =16GB.

Software Requirements:
• Operating System: Windows XP/7/8/8.1/10, Linux and Mac
• Programming Language: Python
• Python Libraries:

1. Pandas
2. Numpy
3. Tensorflow
4. Scikit Learn
 References

[1] K. Gnana Sandhya, S. Vemuri, K. Sai Deeksha and T. Anvitha, "Crop Recommendation System Using
Ensembling Technique," 2022 International Conference on Breakthrough in Heuristics And Reciprocation of
Advanced Technologies (BHARAT), Visakhapatnam, India

[2] D. N. V. S. L. S. Indira, M. Sobhana, A. H. L. Swaroop, V Phani Kumar, “KRISHI RAKSHAN - A Machine

Learning based New Recommendation System to the Farmer”, presented at the 2022 6th International Conference
on Intelligent Computing and Control Systems (ICICCS), Madurai, India

[3] C. Sagana et al., "Machine Learning-Based Crop Recommendations for Precision Farming to Maximize Crop
Yields," 2023 International Conference on Computer Communication and Informatics (ICCCI), Coimbatore, India

Computer Engineering Department 20

Conclusion
➢ This study proposed a system that represents an advancement in optimizing agricultural
practices through machine learning and predictive analytics.

➢ The proposed model has been successfully developed to accurately predict suitable crops
and fertilizers based on a rich dataset comprising features for better recommendations.

➢ The research involved data preprocessing, including cleaning, feature engineering, and
normalization, which ensured the reliability of the results.

➢ sustainable agricultural resource management, offering substantial benefits to farmers and

the global food supply chain.
Thank You…!