Towards an Efficient Deep Learning Approach for Crop

Recommendation

D. Manoj Kumar and S. Sheeja

Karpagam Academy of Higher Education, Coimbatore, India

Keywords: KNN, Random Forest, Deep Learning, Decision Tree, Crop Recommendation.

Abstract: The agriculture industry depends heavily on crop output, which is influenced by a variety of meteorological

and chemical elements. Just because of these two factors, farmers lose a lot of money. Although the climatic

factors are beyond human control, automated technologies may be used to control the chemical factors. Many

solutions to these worries have been offered through research. This study, however, focuses on crop

suggestions based on chemical and meteorological conditions. In order to recommend better crops depending

on chemical and meteorological circumstances, this study proposes an optimization-based deep learning

approach based on the grey wolf. According on a number of chemical and meteorological factors, such as pH,

nitrogen, phosphorus, and potassium as well as rainfall, temperature, and humidity, this paper makes crop

recommendations to farmers. The entire plan is presented in layers: A high-performance Convolution neural

network is utilized to extract and categories important characteristics, after which the feature is optimized

using the grey wolf method to suggest a better crop based on a variety of criteria.

1 INTRODUCTION

Agriculture has historically been and still is one of the

main pillars of the Indian economy because it directly

supports two thirds of the country’s population.

Furthermore, significant is the fact that it accounts for

20% of India's GDP (GDP). The farmer, who serves

as our nation's Annadatta (Food Supplier), is at the

centre of the agricultural industry and is currently

dealing with a number of challenges: 1) Due to the

wide variety of soil types in the nation, farmers

frequently struggle to select the crop that is most

lucrative for their soil, environmental conditions, and

geographic region and consequently suffer significant

losses. 2) Farmers currently find it incredibly

challenging to predict the yield for a particular

sowing season and the profit they can achieve

because to the fluctuating weather conditions. The

depressingly low returns that farmers obtain for their

production are a result of the "farm to market" system,

which is made up of hundreds of intermediaries that

squander the majority of the revenues by transporting

and selling items (Devdatta et al 2019, Anguraj.Ka et

al. 2019). Artificial intelligence, deep learning, and

machine learning are widely used in modern

agriculture (Manish Kumar et al 2022). The general

quality of the harvest, yield forecast, plant pest

detection, and undernutrition of farms can all be

improved with the help of methods like precision

agriculture and crop recommender systems. AI

system deployment could provide the struggling

agricultural industry a boost.

2 LITERATURE REVIEW

Prediction of Crop Yield and Fertilizer

Recommendation Using Machine Learning

Algorithms

Crop yield analysis is a developing study area that

includes machine learning. A major problem in

agriculture is yield prediction. Any farmer is curious

to know how much of a crop he can anticipate. In the

past, farmer experience with a certain field and crop

was taken into account when predicting production.

Based on the information at hand, the yield forecast

is a significant problem that needs to be resolved. The

more effective option for this is machine learning. In

order to predict the crop yield for the next year, many

machine learning algorithms are utilised and tested in

agriculture (Devdatta et al 2019). In this study, a

technique to forecast agricultural yield using

historical data is proposed and put into practice. This

is performed by applying machine learning

Kumar, D. and Sheeja, S.

Towards an Efﬁcient Deep Learning Approach for Crop Recommendation.

DOI: 10.5220/0012614100003739

Paper published under CC license (CC BY-NC-ND 4.0)

In Proceedings of the 1st International Conference on Artiﬁcial Intelligence for Internet of Things: Accelerating Innovation in Industry and Consumer Electronics (AI4IoT 2023), pages 281-285

ISBN: 978-989-758-661-3

281

algorithms to agricultural data, such as Support

Vector Machine and Random Forest, and making

fertilizer recommendations that are suitable for each

unique crop. The research focuses on developing a

prediction model that could be applied to crop yield

forecasting in the future. It gives a succinct

description of how agricultural yield can be predicted

using machine learning methods.

Figure 1: Flow chart of Proposed Model.

Crop Recommendation on Analyzing Soil Using

Machine Learning

Influence on crop productivity. By implementing

emerging Agriculture into the effect, our

government's crops are stretched, which further

greatly boosts the economy of our nation. Crop

productivity has been strongly influenced by

fluctuations in the weather. Modern paradigm has the

potential to replace conventional farming with

precision farming, that will further improve

agricultural yields. Two examples of contemporary

technology in use are preliminary study and indeed

the internet of things (IOT). The fundamental issue

that remains to be investigated is growing precisely

crops at precise times (Anguraj.Ka et al. 2019). This

can be done using machine learning techniques,

which have been found to be an effective strategy for

predicting the perfect crop. Agriculture IOT sensors

are used to gather the soil data, such as soil moisture,

temperature, humidity, and pH, and provide them to a

graphical user interface (GUI). The GUI gathers the

inputs and recommends the appropriate crops. The

system created with IOT and ML significantly aids

farmers in making wise decisions.

Crop Recommendation Predictor Using Machine

Learning for Big Data

The needs of a constantly expanding human

population. In rural areas of India, agriculture is the

primary industry. We all know that the majority of

Indians work mostly in agriculture. Most Indians,

either explicitly or implicitly, depend on agriculture

for their livelihood. Most farmers in India rely on

their instincts to choose which crop to grow during a

certain season. Farmers are used to planting the same

crop, applying additional fertiliser, and adhering to

popular opinion. They are unaware of the fact that

crop productivity is heavily reliant on the current

weather, soil, and other factors and instead find

comfort in just adhering to previous agricultural

traditions and norms. The most frequent issue Indian

farmers have is that they don't choose the right crop

depending on the needs of their soil and other

elements like fertilisers and irrigation schedules

(Manish Kumar et al 2022. Nikam et al 2022).

Productivity suffers as a result. A single farmer,

however, cannot be expected to take into account all

of the various factors that affect crop development

when determining which crop to plant. Machine

learning is an effective solution to this issue.

Throughout the past few years, significant

improvements have been made in the ways that

machine learning can be used in numerous studies and

enterprises. As a result, we want to create a model or

system that will let farmers use machine learning in

agriculture.

3 PROBLEM DEFINITION

Few platforms exist that aid farmers in developing

their farming strategies. Decisions based on intuition

might not turn out to be advantageous in the long run.

Farmers frequently overestimate or underestimate the

soil fertility on their fields. They frequently have

trouble identifying plant illnesses that have an

immediate impact on the rate of output. It is feasible

to provide precise crop forecast results by using the

right criteria, such as rain patterns, temperature

patterns, soil structures, and other things like crop

diseases. Furthermore, it is also feasible to determine

in advance what disease a crop has. Several of the

current systems are exceedingly difficult to use or

have numerous problems that make them unintuitive.

AI4IoT 2023 - First International Conference on Artiﬁcial Intelligence for Internet of things (AI4IOT): Accelerating Innovation in Industry

and Consumer Electronics

282

Figure 2: (a) Temperature; (b) Ph; (c) Crops.

Figure 3: (a) Humidity; (b) Ph.

Figure 4: (a) Rainfall; (b) Humidity.

Towards an Efﬁcient Deep Learning Approach for Crop Recommendation

283

Figure 5: (a) Correlation Between different features; (b) Temperature and crops; (c) Confusion Matrix.

4 PROPOSED MODEL

India's agriculture sector is significant. It is necessary

for the Indian economy's survival and expansion.

India is a significant producer of many different

agricultural goods. In the process of cultivating crops,

soil is crucial. A non-renewable, dynamic natural

resource required for life is soil. Crop cultivation used

to be done by farmers with practical experience.

Based on the qualities and properties of the soil,

farmers are no longer able to select the ideal crop.

Hence, a recommendation system that uses machine

learning algorithms to suggest the crop that can be

harvested in that specific soil has been developed. To

advise the crop, this system employs a variety of

machine learning techniques, including KNN,

Decision Tree, Random Forest, Naive Bays, and

Gradient Boosting.

5 RESULTS AND DISCUSSIONS

The proposed model, designed with the objective of

enhancing crop recommendation in agriculture using

machine learning, was implemented using the Python

programming language. The methodology

encompassed a multi-stage approach involving data

extraction, feature engineering, model training, and

evaluation. The dataset incorporated numerous

chemical and meteorological factors, such as pH

levels, nitrogen, phosphorus, potassium, rainfall,

temperature, and humidity. These variables were

carefully selected due to their significant impact on

crop growth and yield. The final iteration of the KNN

architecture exhibited an impressive accuracy rate of

91.21%. This high accuracy underscores the efficacy

of this algorithm in suggesting suitable crops based

on the given parameters. The Random Forest

Algorithm, while slightly lower in accuracy

compared to KNN, achieved a respectable accuracy

rate of 75%. Further fine-tuning and optimization

might be required to enhance its performance.

6 CONCLUSIONS

The agriculture sector's health is essential for India's

long-term economic success. By boosting

profitability and enhancing agricultural productivity,

our goal was to give small-scale farmers more

control. In our trials, a variety of machine learning

techniques are used to recommend the crop, including

AI4IoT 2023 - First International Conference on Artiﬁcial Intelligence for Internet of things (AI4IOT): Accelerating Innovation in Industry

and Consumer Electronics

284

KNN, Decision Tree, Random Forest, Naive Bays,

and Gradient Boosting. The astounding accuracy of

the KNN architecture's final version was 91.21%. The

accuracy of the Random Forest Algorithm was 75%.

To further enhance the system and provide more

precise yield prediction findings, the crop production

dataset can be expanded.

REFERENCES

Devdatta A. Bondre, Mr. Santosh Mahagaonkar (2019)

“Prediction of Crop Yield and Fertilizer

Recommendation Using Machine Learning

Algorithms” International Journal of Engineering

Applied Sciences and Technology, 2019Vol. 4, Issue 5,

ISSN No. 2455-2143, Pages 371-376Published Online

September 2019 in IJEAST.

Anguraj.Ka, Thiyaneswaran.Bb, Megashree.Gc,

Preetha Shri.J.Gd, Navya.Se, Jayanthi.(2019) Jf “Crop

Recommendation on Analyzing Soil Using Machine

Learning” Turkish Journal of Computer and

Mathematics EducationVol.12 No.6 (2021), 1784-

1791.

Manish Kumar, Deva Prakash (2022) “Crop

Recommendation Predictor using Machine Learning

for Big Data” International Journal of Engineering

Research in Computer Science and Engineering

(IJERCSE)Vol 9, Issue 8, August 2022.

Nikam D. A., Amanul Rahiman Shamsuddin Attar, Omkar

Yashwant Bujare, AkilAppasoBandgar, Shridhar

Rajaram Banne “Intelligent Crop and Pesticide

Recommendation Portal Using ML and AI”

International Research Journal of Modernization in

Engineering Technology and Science (Peer-Reviewed,

Open Access, Fully Refereed International Journal)

Volume:04/Issue:06/June-2022.

Towards an Efﬁcient Deep Learning Approach for Crop Recommendation

285