National E-Conference on "Communication, Computation, Control and Automation" ( CCCA-2020)

Available at www.ijsred.com
RESEARCH ARTICLE OPEN ACCESS

Soil Nutrient Analysis Using Machine Learning Techniques

Madhumathi R1,Arumuganathan T2, Sneha Iyer R3, Shruthi R4, Shruthhi K5
1,3,4,5
Department of Computer Science and Engineering,
Sri Ramakrishna Engineering College, Coimbatore, India
Email:[email protected] ,[email protected]
2
Principal Scientist,Sugarcane Breeding Institute, Coimbatore, India
Email:[email protected]
-----------------------------****************************-----------------------------
Abstract:Agriculture is an important aspect in determining the overall growth of a nation, both in terms of
economic and also in terms of progress in the people’s lives. Nutrient analysis forms the main criteria for
producing a healthy crop. Hence,adopting different measures for soil analysis has become the need of the hour.
Primarily, thesoil analysis prediction consists of determining the composition of nitrogen(N),phosphorus(P) and
the potassium(K)content present in the soil. These are the major nutrients that the crop requires for its growth.
Machine learning algorithms are very effective in predictive analytics and are widely used in the field of
agriculture.Ithasa set of well-defined models like classification, regression, Support vector machine(SVM)or
clustering that collect specific data and apply specific algorithms to achieve expected results. The system aims at
the providing a feasible solution which can be obtained through the machine learning model. For determining the
N, P, K composition in the soil the multiple linear regression(MLR) algorithms is used. Thus, this method would
help the farmers in choosing the right crop and fertilizer to have better harvest of crops. The prediction technique
of soil nutrient contentwould thereby increase the country’s overall economy among other nations.

Keywords-Multiple linear regression,Soil nutrients,Nitrogen,Phosphorus,Potassium.

---------------------------******************************----------------------------
I.INTRODUCTION agriculture. Smart farming powered by machine
Agriculture is a vital aspect for a nation in learning with its high-precision algorithms is an
determining its overall progress among the other emerging concept. It includes crop quality
nations. It is the art of living in the country like identification using image processing, yield
India.Even with the establishment of the prediction using various algorithms, disease
technological era,agriculture still continues to be the detection using sensors and algorithms, field
one of the major areas wherein technical conditions monitoring like soil management, water
advancements are highly appreciated. To have an management and crop management. There are
effective production of the crop and adding various algorithms that are used for addressing
fertilizers in the right ratio in the soil, it is important different algorithms. Machine learning algorithms
for the farmers to know the soil nutrient like regression, classification aremainly used for
composition. Hence, determination of soil nutrient prediction and it proves to give valuable results.
analysis has become the need of this today’s There various other techniques as well that are used
world.Themachine learning algorithm plays a major in agricultural field. Sensor, devices with smart
role and gives faster and accurate results.The optimization techniques are also employed in this
quantity of the N, P, K present in soilcan be field. The main motto is to predict the results
determined using the laboratory methods. Here the accurately through the implementation of the MLR
machine learning algorithm is used for the technique
prediction for soil nutrients.
It is a booming technology in the field of

ISSN : 2581-7175 ©IJSRED:All Rights are Reserved Page 1

 National E-Conference on "Communication, Computation, Control and Automation" ( CCCA-2020)
Available at www.ijsred.com
The artificial neural network(ANN)model humidity and N, P, K values are displayed in a
was developed for determining the soil nutrients for mobile application. The soil pH, Electrical
a specific plant [1]. Machine learning algorithms Conductivity, Moisture, temperature and N, P, K
like multiple linear regression and classification values were analyzed and based on the data obtained
algorithms were used for comparingtheir results in the accurate crop to be grown in that soil along with
soil nutrient analysis[2].A study was proposed the suggestion of fertilizers were predicted using
which focuses on the various types of the soil,crop Supervised Machine learning algorithms[10].
types and the soil test reports. Asuitable machine
learning model was developed for prediction using A study was proposed where the soil features
the implementation of the k-nearest like village wise soil fertility indices of Available
neighbor’stechnique[3].The determination of the soil Phosphorous, Available Potassium (K), Organic
nutrient analysis using recurrent neural Carbon (OC), Boron (B) and Soil Reaction(pH) are
network,SVM and MLRmodels were developed. classified by the index analysis test report
The accuracy of each model was compared[4]. values[11]. The classification is done using the
Extreme Learning Machine (ELM) with different
Thispaperdiscussesthemethodologyandsoluti activation functions. This learning helps in the better
onemployed to solve the problem for detecting soil production of better results.A system was designed
nutrients. Henceforth, it would be very useful for the where the Soil fertility index for the nutrients is
farmers for growing better crops. predicted at different places. The prediction is done
automatically using regression methods specifically
II.RELATED WORK like support vector regression and more[12].
The soil nutrients were analyzed using the
regression algorithm and the root mean square error The data obtained is useful for making
is calculated for the prediction of N, P and K[5]. A decision of amount of fertilizers to be used. A
consecutive comprehensive evaluation models has system was proposed to determine the pH value of
been presentedfor the soil nutrient analysisusing the soil and based on the pH value obtained the NPK
SVM, ANN and MLR[6]. A study has been values are estimated[13]. The SVM algorithm and
conducted on the various machine learning the artificial neural networks play a major role in the
techniques which are available for crop analysis of soil nutrient[14].Using classification
determination and other agriculture related algorithms, the suitable crops and the amount of
practices[7].The soil fertility and the plant nutrients fertilizer is recommended for that land in a handheld
were analyzed using Supervised Machine learning, device.
Back Propagation Neural Network(BPN)[8].The
reference crop’s growth properties are trained to find III.PROPOSED SYSTEM
the actual values. BPN will find the correlation The proposed system predicts the amount of soil
percentage of the properties trained and its nutrients namely N,P,K using the trained dataset
performance is evaluated using the test data. A
which has various parameters and the multiple linear
system was proposed which integrates a sensing
regression algorithm. Thus, knowing these values of
module with an image processing setup to determine
the essential characteristics for plant growth. The soil nutrients, it would be easier for the farmers for
image processing mainly helps in the capturing the adding fertilizer in a right range and produce better
soil images. Features extraction is performed for yield.
determining the results[9].The feature values A.Methodology
obtained were analyzed by comparison with the The aim is to predicted the values of N,P,K
database.SVM was used to classify the ratio level of using the machine learning algorithm. The dataset is
N, P, K. The soil moisture, pH, temperature,
ISSN : 2581-7175 ©IJSRED:All Rights are Reserved Page 2
 National E-Conference on "Communication, Computation, Control and Automation" ( CCCA-2020)
Available at www.ijsred.com
being imported in RStudio platform, it consists of Figure.1 is the block diagram of the
parameters like the electrical conductivity, pH, soil proposed system which shows the process of
moisture, soiltemperature and the corresponding predicting the nutrient values. It starts from
values of N,P,K in the soil. These parameters play a importing of the dataset that has all the
significant role in development and growth of a plant. parameters and the respective N,P,K
Thus, these parameters are used as variableswhere the values.Train the model with the multiple linear
regression algorithm. The model is being tested
nitrogen, potassium and phosphorous forms the target
with different values as input to check its
variable respectively.
efficiency. The graph is being plotted for the
The dataset is trained using the MLR model. respective target values by taking predicted
This model is used because of its efficiency to handle values as x-axis and the parameters as y-axis.
multiple variables. The scatterplot and pairs function
which is used for the creation of graph of the model B. RStudio
that shows the contribution of each parameter. The RStudio is a platform that is an
There are different values of each nitrogen, integrated development environment for R that
allows to perform various functions related to
potassium and phosphorous values predicted after
the predictive,statistical and graphical analysis.
giving the parameters pH, moisture, electrical
Various packages can be used for different
conductivity, temperature as an input. The predicted applications and its implementation.
values are then used for creating graphs for N, P, K Thus,RStudio is a great platform for performing
respectively. machine learning algorithms, predicting values
and for graphical analysis.
Start
End C. Multiple Linear Regression Algorithm
It is the most common form of Linear
Regression. It basically describes how a single
Import the
response variable Y depends linearly on a
dataset Plot the result number of predictor variables. The model is
in a graph
mainly used when there are multiple parameters
and one target variable to be found.
Yes
The target variable is the one that would
be depended on the other parameters.Thus,
Train the model
using Multiple Test the thevariables like the pH, electrical conductivity,
Linear Regression model
soil moisture and soil temperature are
considered. The dataset is trained to get the
predicted values of N,P,K. The general formula
of multiple linear regression algorithm:
If not accurate Y = βo + β1x1 + …… +βkxk + e
repeat the process to
get the accurate No
model

Fig.1 Block Diagram of the Proposed System

ISSN : 2581-7175 ©IJSRED:All Rights are Reserved Page 3

 National E-Conference on "Communication, Computation, Control and Automation" ( CCCA-2020)
Available at www.ijsred.com
IV.RESULT ANALYSIS TABLE 4
THE PREDICTED VALUES OF POTASSIUM
The table1 shows that the dataset’s Temperature Moisture pH Conductivity Potassium
variables have huge significance the highest one 24 90 4.7 0.025 16.11
goes to the moisture [***] and then the pH [**] 17 68 4.1 0.05 21.54
has its second most significance and then comes 25 98 4.2 0.35 8.38
the conductivity [**] which has third most 24 92 5.2 0.31 9.72

significance. It has a p-value of 2.847e-09 which 19 75 4.2 0.28 14.96

tells that the model is quite efficient and thus Thus, the table2 shows the predicted value of
most of the data contributes for the prediction. nitrogen when the other parameters temperature,
moisture, pH and conductivity given as input.
TABLE 1 Thetable3 shows predicted value of potassium when
CONTRIBUTION OF EACH VARIABLE
the other parameters like the temperature,
moisture,pHand conductivity given as
Estimated standard error t value pr(<|t|)
input.Thetable4shows the predicted the value of
Intercept 530.009 161.064 3.291 0.00141 **
phosphorous aswhen the other parameters
temperature, moisture,pH and conductivity given as
Temperature 5.737 3.364 1.706 0.09137 * input.Henceforth, the predicted values could be
compared with the dataset values to the efficiency of
Moisture 6.826 1.36 5.02 2.44-e06 *** the model.
pH 33.796 10.483 3.224 0.00174 **

Conductivity 338.417 127.663 2.651 0.00942 **

The amount of nitrogen,potassium and

phosphorus is predicted using the multiple linear
regression and gives an accuracy of the model for
about 78%.
TABLE 2
THE PREDICTED VALUES OF NITROGEN

Temperature Moisture pH Conductivity Nitrogen

24 60 5 0.045 320.27
17 85 4.7 0.035 312.41
19 75 5 0.035 329.92
15 95 5.2 0.25 381.63 Fig.1Predicted graph of Nitrogen
25 55 4 0.031 310.16

TABLE 3
THE PREDICTED VALUES OF PHOSPHOROUS
Temperature Moisture pH Conductivity Phosphorous
25 99 5 0.31 353.25
21 84 5.1 0.36 214.37
24 93 5.2 0.32 309.94
18 73 5.1 0.34 128.85
20 83 4 0.24 205.25

ISSN : 2581-7175 ©IJSRED:All Rights are Reserved Page 4

 National E-Conference on "Communication, Computation, Control and Automation" ( CCCA-2020)
Available at www.ijsred.com
V. CONCLUSION
This system predicts the values of N,P,K
using the multiple linear regression model and gives
the composition of the soil nutrient i.e. N, P, K
content present in the soil. The multiple linear
regression model used for prediction gives an
accuracy of 78%. This accuracy makes it almost a
suitable approach as well as a trustworthy technique
in determining the fertility of the crop with a good
amount of efficiency maximum possible. Hence
machine learning technique proves to be an effective
method for prediction of soil nutrient and as well as
its analysis.
Thiscould further be extended by developing
a mobile application where the values of N, P,K that
Fig.2 Predicted graph of Phosphorous is obtained from the model could be displayed in the
application developed. The values can be stored in a
database and displayed in the app. Hence the system
benefits the farmers and helps them taking right
decisions for growing crops.

ACKNOWLEDGMENT
The authors sincerely thank the Science and
Engineering Research Board (SERB)for the
financial support, the Director, ICAR-SBI,
Coimbatore, and the Management and Principal of
Sri Ramakrishna Engineering College, Coimbatore
for extending the required facilities for this study.
REFERENCES
[1] C.Wu, etal “Evaluating soil nutrients of DacrydiumpectinatuminChina
using machine learning techniques”, InternationalJournal of forest
Fig.3 Predicted graph of Potassium ecosystems, Volume 6, Issue 7,2020.

[2] G.Forkuor et al “High resolution mapping of soil properties using remote

Hence these are the predicted values of N,P,K sensing variables”, Plos one publishers, 2017.
where the line red depicts moisture, black [3] P.Yadav,“Soil health analysis for crop suggestions using machine
depicts temperature,bluedepictspH and orange learning”, International Journal of research and development, Volume 6,
Issue 12, 2019.
conductivity. The fig.1 shows the graph of
predicted values of nitrogen to the parameters. [4] S.Prakash, “Soil moisture predictionusing machine learning”, Institute of
electrical electronics engineering, Volume 8, Issue 7,2018.
The fig.2 shows the graph of predicted values of
phosphorous to the parameters. The fig.3 shows [5] K.Kumar, “Random forest algorithm for soil fertility prediction and
grading using machine learning “International journal of innovative
the predicted values of potassium to the technology and exploring engineering, Volume 1, Issue 9,2019.
parameters. [6] H.Li, “Evaluation models for soil nutrient based on support vector
machine and artificial neural networks”,Hindawi publishers,2014.

[7] Konstantinos and etal, “Machine learning in agriculture”,

InternationalJournal of Innovative Science and Research, Volume 6,

ISSN : 2581-7175 ©IJSRED:All Rights are Reserved Page 5

 National E-Conference on "Communication, Computation, Control and Automation" ( CCCA-2020)
Available at www.ijsred.com
Issue 7,2018.

[8] S.Ghosh and S.Koley, “Machine Learning for Soil Fertility and Plant
Nutrient Management using Back Propagation Neural Networks”,
International Journal on Recent and Innovation Trends in Computing and
Communication(IJRITCC), Volume 2, Issue 2, pp.292-297, 2014.

[9] S.Rahaman, et al “Detection of NPK Ratio Level Using SVM Algorithm

and Smart Argo Sensor System”, International Journal of Latest Research
in Engineering and Technology (IJLRET), Volume 3, Issue 7, pp.11-15,
2017.

[10] S. Panchamurthi, M. D. Perarulalan, A. Syed Hameeduddin and P.

Yuvaraj, “Soil Analysis and Prediction of Suitable Crop for Agriculture
using Machine Learning”, International Journal for Research in Applied
Science & Engineering Technology (IJRASET), Volume 7, Issue 3, pp.
2328-2335, 2019.

[11] M.S. Suchitra and Maya L. Pai,“Improving the prediction accuracy of

soil nutrient classification by optimizing extreme learning machine
parameters”, Information Processing in Agriculture, Elsevier, 2019.

[12] M.S. Sirsat, E. Cernadas, M. Fernández-Delgado and S. Barro,

“Automatic prediction of village-wise soil fertility for several nutrients in
India using a wide range of regression methods”, 2018, Computers and
Electronics in Agriculture, Volume 154, pp.120-133.

[13] K.Abhang, S.Chaughule, P.Chavan and S.Ganjave, “Soil Analysis and

Crop Fertility Prediction”, International Research Journal of Engineering
and Technology (IRJET), 2018, Volume 5, Issue 3, 2018.

[14] J.C Puno, E.Sybingo, E.Dadios, I.Valenzuela and J.Cuello,

“Determination of Soil Nutrients and pH level using Image processing
and Artificial neural networks”, IEEE,2017.

ISSN : 2581-7175 ©IJSRED:All Rights are Reserved Page 6