Crop Yield Prediction

Abhishek Sahani1, Deepak Kumar Mishra2, Sumit Kumar Gupta3, Livingston Jeeva4
1,2,3
Student, Department of Computer Science and Engineering, School of Management Sciences Lucknow
4
Assistant Professor, Department of Computer Science and Engineering, School of Management Sciences Lucknow
Corresponding Author: abhisheksahani487@gmail.com

Abstract: Everyone in India is aware that agriculture is the foundation of our nation. This article forecasts the output of
nearly every crop cultivated or farmed in India. Using basic variables like State, District, Season, and Area, the user of this
script can forecast the crop's yield for the year they wish to grow or produce it. The Random Forest Algorithm (RFA) of
Supervised Learning and the idea of regression to enhance the algorithms are two examples of sophisticated regression
approaches used in the study to forecast the yield. While there is still much to be done in many disciplines, technological
breakthroughs have already shown improvements in numerous areas. It will show the ratio of the soil's phosphorus,
potassium, and nitrogen levels.
Key Word: Farming, Machine Learning Techniques, Indian Agriculture, Random Forest Algorithm, Proper Yield

I. Introduction
In our research work, we found that all climate factors like soil type, rainfall, solar nutrients (nitrogen,
phosphorus, etc.) are used by humans, but the problem is that the data needs to be collected. A third party makes
predictions and explains them to the farmer. Farmers and growers who do not understand the science behind these methods
require a lot of time and effort. In this study, we use simple variables like farmer's personality, residential area, crops and
weather to ensure that it is fair and suitable for farmers. In India, more than hundreds of fruits are grown all over our
country. The cut is broken for better understanding and visualization. The data for this study was obtained from the
archives of the Government of India. With around 4.5 lakh views, the data includes many parameters including state,
region, product, season, year, location and production. Many states and regions in India feature the most recognizable
product categories and seasons. To reduce errors and generate accurate predictions for different types of products, we use a
forward regression approach, including a Random Forest Algorithm (RFA) based on supervised learning. It is the largest
and most important community in India. Because ancient peoples grew food on their own agricultural lands and could meet
their own needs. As a result, natural foods are grown and used by many species, including humans, animals and birds.
Living things enjoy life and prosperity by eating the green things produced from the soil. The agricultural sector is
deteriorating since the appearance of new technologies and methods.

II. Literature Survey

Crop yield prediction studies are conducted using a variety of machine learning techniques and algorithms
available to agricultural institutions. This paper reviews some of these studies conducted in research papers with the
techniques used and results.
1. Anantara, M.G. et al. (February 2013) In "Survey On Crop Yield Prediction", a predictive model for agricultural
datasets called his SVM algorithm for crop yield using regenerative techniques is proposed. They considered
parameters such as plant species, soil and soil pH, humidity, and plant sensitivity. Their analysis was mainly
concerned with wet rice and sugarcane yields in India. The proposed algorithm is a support vector machine SVM
algorithm, which performs well with an accuracy of 85%.
2. Awan, A.M. et al. (April 2006) “Survey on Crop Yield Prediction”, focuses on the clustering decision tree method
for predicting crop yields and analyzes parameters such as plantation, temperature, and rainfall. We've created a
new, intelligent framework that takes you into account. their different regions. Experimental weighting of
decision tree methodology with spatial constraints for analysis of palm fields.
3. Chawla, L et al. (August 2019) “Survey on Crop Yield Prediction”, Fuzzy logic was used to predict crop yield
using a statistical time series model. They considered various parameters
4. Chaudhari, A.N. et al. (August 2018) “Online Survey on Crop Yield Prediction”, two algorithms are used to
improve crop yield prediction: clustering K-means and bias algorithm. I have used many other algorithms to use it
effectively. It was also found that parameters such as pH levels of different regions, rainfall, and soil types were
also taken into account, allowing the system to identify which crops are highly suitable for cultivation based on
the aforementioned characteristics.

1 | Page
 5. Manjula, A. et al. “Agro-meteorological yield prediction using machine learning”, create crop selections and
predict yields by considering various indicators such as vegetation, temperature, humidity, and various vegetation
factors. Then, to improve understanding, climatic and agricultural factors and other disturbances caused by
projections are distinguished.
6. Tripathy, A. K. et al. (July 2016) “Survey On Crop Yield Prediction”, Some of the authors use support vector
machine (SVM) algorithm to predict high crop yield with the same feature. Predicted yield, states that a novel
decision-making system is proposed that provides inputs and an interface to achieve better results when applying
SVM and neural networks to predict high crop yields. We also considered climatic factors such as temperature,
precipitation, and evapotranspiration, primarily (taking into account the range) of the reference plants.
7. Ishwar Chawla et al. Forecasting Temperature and Rainfall Factors Using ARMA, SARIMA, ARMAX Models,
Fuzzy Logic Based Crop Yield Forecasting: Agriculture is of great importance in the Indian economy. Predicting
crop yields is critical to India's prosperity. Various weather factors such as temperature and precipitation affect
crops. Therefore, it is important to consider these aspects when estimating crop yield. It's difficult to predict the
weather. In this study, ARMA, SARIMA, and ARMAX (autoregressive integrated moving average) are used as
three forecasting methods (ARMA with exogenous variables). The results of the three models are compared and
the most accurate model is used to predict temperature and precipitation, which are then used to predict crop
yields.

III. Methodology
The algorithms used in our project are Random Forest Algorithm (RFA) and Neural Network. The measured results of our
experiments show that the RFA algorithm outperforms other algorithms and not only achieves the highest accuracy in
weather forecasting, but also helps in achieving high crop yields. Depending on the soil quality, the system estimates crop
yield and recommends fertilizers if necessary. The results are processed by two controllers. The address is used as input to
the controller in addition to using third-party applications such as APIs for weather and temperature, soil type, local soil
nutritional value, local precipitation, soil composition, etc. calculated. It then takes the pH value as input and uses that
value to determine the alkalinity of the soil. Also, the proportion of nutrients such as nitrogen (N), phosphorus (P), and
potassium (K).

Neural Networks
Neural networks, a subset of machine learning, often referred to as artificial neural networks (ANNs) or simulated neural
networks (SNNs), are the basis of deep learning approaches. Their organization and terminology are modeled after the
human brain and mimic the communication of real neurons. The input layer, one or more hidden layers, and the output
layer form the node layers of an artificial neural network (ANN). Each node or artificial neuron is connected to other nodes
and has weights and thresholds. A node whose output exceeds a specified threshold becomes active and begins sending
information to the top layer of the network.

Figure- I [Neural Network]

Random Forest Algorithm

2 | Page
Random Forest is a technique used in supervised learning methodologies. You can use it to solve ML problems involving
both classification and regression. It is based on the concept of ensemble learning, which is a way to combine different
classifiers to tackle difficult problems and improve model performance. Random Forest, as the name suggests, is a
classifier that takes a series of decision trees on different subsets of a provided dataset and averages them to improve the
predictive accuracy of the dataset. Rather than relying on a single decision tree, a random tree uses predictions from each
tree and predicts the outcome based on a large proportion of the predicted votes.

Figure-2 [Random Forest Algorithm]

Existing Systems and Disadvantages

The current system does not have a computerized system for identifying recommended culture or behavioral analysis.
First, it is only suitable for instance-level techniques that require an instance classifier. As mentioned previously, current
common neural network techniques take discrete instances as input and transform them into an embedding space using a
deep neural network.

Advanced Systems
Proposed Method as environmental variables vary from region to region, a machine learning model is used to estimate the
ideal plant variety for a selected plot. It uses data from standard datasets to train crop suggestion models and uses machine
learning techniques to select the best crops with the highest growth probability for cultivation. Naive bias and random
forestry machine technology support determines the optimal crop type. Based on this model, it was determined which types
of crops farmers should grow. Aspects to consider include humidity, temperature, soil moisture, and rainfall.

Advantages
1. Efficient and safe system.
2. The advantage of this approach to farming is that it allows selection of appropriate crops based on factors such as
climate and temperature.
3. Machine learning helps avoid assumptions, challenges in using larger sample sizes, and complications and makes
predictions based on available data.

IV. Conclusion
This project leverages machine learning and uses random forests, polynomial regression, and decision tree techniques to
evaluate performance. Random Forest provides the best yield prediction among the three algorithms of the proposed model
compared to other methods. The output showing improvements in the dataset is classified using random forest regression
techniques and neural network models. After the analysis, we concluded that the proposed model is more effective than
current approaches to determine crop yield. Adopting the above methods will help in improving the agricultural practices
of our country. Additionally, it can also be used to reduce farmers' losses and increase crop yields, thereby increasing
agricultural capital. Models can be improved through integration. Using stacked regression results in less predictable
results than using each model individually.

References
[1]. Paul, M., S. K. Vishwakarma, and A. Verma (2015, December). Data mining approach for soil behavior analysis and crop yield prediction.
2015's Computational Intelligence and Communication Networks International Conference (CICN) (pp. 766-771). IEEE.

3 | Page
[2]. R. Jaikla, S. Auephanwiriyakul, & A. Jintrawet (2008, May). Using support vector regression, estimate rice yield. The Fifth International
Conference on Electrical Engineering/Electronics, Computer, Telecommunication, and Information Technology was held in 2008.
[3]. Ramalatha, M., Suresh, A., and Kumar, P. G. (2018, October). Tamil Nadu's main agricultural yields are predicted using K-means and
Modified KNN. The third International Conference on Communication and Electronics Systems (ICCES) was held in 2018. (pp. 88-93). IEEE.
[4]. Ananthara, M. G., Hemavathy, R., & Arunkumar, T. (2013, February). CRY is a more accurate model for predicting crop yields from
agricultural data sets utilizing bee hive clustering. In 2013 International Conference on Informatics, Mobile Engineering and Pattern
Recognition.
[5]. KM Meheta (January 2015). XCYPF: A adaptable and scalable framework for crop yield prediction in agriculture. 9th International
Conference on Intelligent Systems and Control (ISCO) of the IEEE was held in 2015.
[6]. Misra, B., Veenadhari, S., and Singh, C. C. (2014, January). A method using machine learning for predicting crop yield based on climate
factors. Global Conference on Computer Communication and Informatics in 2014.
[7]. S. Bang, R. Bishnoi, A. S. Chauhan, A. K. Dixit, & I. Chawla (2019, August). Crop Yield Prediction Based on Fuzzy Logic Using Predicted
Temperature and Rainfall Parameters from ARMA, SAR/MA, and ARMAX Models. The 12th International Conference on Contemporary
Computing (IC3) will be held in 2019.
[8]. Shah A, Dubey A, Hemnani V, Gala D, and Kalbande D. R. (2018). Crop Yield Prediction Using Regression Techniques in a Smart Farming
System. International Conference on Wireless Communication Proceedings (pp. 49-56). Singapore's Springer.
[9]. Sap, M. N. M., and Awan, A. M. (2006, April). a kernel-based intelligent system for forecasting crop yields, Conference on knowledge
Discovery and Data Mining in Pacific-Asia.
[10]. Miniappan, N., Kadir, M. K. A., and Ayob, M. Z. (August 2014). Prediction of wheat yield using artificial neural networks. The fourth
International Conference on Engineering, Technology, and Entrepreneurship (ICE2T) was held in 2014.

4 | Page