Digital Soil Mapping using Machine Learning

© 2022, IRJET | Impact Factor value: 7.529 | ISO 9001:2008 Certified Journal | Page 2257
Digital Soil Mapping using Machine Learning
Ashu Bansal, Riya Jain, Muskan Rastogi, Vimal Kumar
Department of Computer Science and Engineering, Meerut Institute of Engineering and Technology, Meerut,Uttar
Pradesh, India
----------------------------------------------------------------------------***-----------------------------------------------------------------
ABSTRACT
Agriculture is a non-technical sector where in technology can be incorporated for the betterment. Soil analysis is a method
to analyse the available plant nutrients in the soil. Soil provides major nutrients to the plants. To create a prediction engine
for most appropriate cropfor a particular soil. It also determines the type of soil and its fertility. This work predicts the
suitable crop and the fertility of a particular soil by analyzing the major and micro nutrients present in the soil. There are
mainly three soil parameters that come into consideration when we have to predict the quality of the soil. This method
suggests the soil fertility and suitable crop for a soil using Machine Learning Techniques.Our result gives the compatible
crop for aparticular soil sample by considering important soil parameters and by applying appropriate Machine Learning
algorithm. Suitable crop for a particular sample is predicted on the basis ofNPK factor, type of soil according to the pH level
and soil fertility on the basis of major and micro nutrients with the maximum accuracy.
Keyword : Nutrients, parameters, Machine Learning, Accuracy.
1 INTRODUCTION
Agriculture is a non-technical sector wherein technology can be used for the efficient management of soil. There is quick
implementation and easy in adoption in theagricultural technology. Usually farmers used traditional method known as
crop mutation after every subsequent crop yield [1]. In many countries, this traditional method is implemented where the
change in crop is done after a loss in yield for cultivating the same crop continuously. This method helps the soil to regain
the nutrients that were consumed by the crop previously and use the remaining nutrients for cultivating the new crop.
Soil fertilityis also maintained by this process. Farmer has to face a loss in yield when they come to know about condition
of soil which is unfit to yield the particular crop [2]. Aboutone financial year is important for a farmerto accept the loss in
yield. Solution to the above stated problem is suggested usingMachine Learning Techniques. There are mainly three soil
parameters that becomes necessary in the prediction of soil fertility and suitable crop for a soil sample.
 Chemical Parameters
 Physical Parameters
 Biological ParametersTable 1 Soil Parameters
Parameters Chemical Description
Physical Electrical
Conductivity
Ions present in thesoil sample are measured by the EC
of soil.
Conductivity of the soil increaseswhen there is a
movement of ions.
Texture Land consists ofvarieties of soil such as Clayey, Sandy,
Layered,
Semi-Layered etc.
pH It is a scale used tospecify the acidic or basic character
of an aqueous solution.
Chemical Sulphur It helps in the production of chlorophyll that isrequired
for the photosynthesis.
International Research Journal of Engineering and Technology (IRJET) e-ISSN: 2395-0056
Volume: 09 Issue: 05 | May 2022 www.irjet.net p-ISSN: 2395-0072

A soil sample contains different types of elements that shows different behaviour with different crops. This table shows
those components that are present in a sample.
There are the demerits of the abovetraditional method:
 Scope for redundancy
 Time lag
 Requires more human force
 Needs more labs
Soil Testing and analysis gives the accurate composition of the soil and the respective compatible crop. It directlybenefits
a large number of the user (farmers or people associated with agriculture and farms). Machine Learning is always
important for large dataset.Machine Learning Techniques make theprocessing flexible and automated usingalgorithms [3].
This project helpful to predict the fertility of soil by using some machine learning algorithms like Decision Tree Classifier,
KNN Classifier and Random Forest Algorithm. Theclassification of soil according to thefertility also can be made easy by
analysing the major and micro nutrients of the soil [4]. It takes less time to predict the fertility of soil than the traditional
system as the machines worked faster and more efficient than the manual system. We can also note the fertility and the
type of the soil very easily and efficiently and within less time. Our aim is to come up with an automatic soil testing system
which not only will analyse the soil samples but also provide acceptable crop information at freeof cost and by consuming
less time. This crop prediction is finished by not just considering the fertility of the soil but also by the type of a soil sample
[5].
Using machine learning ideas, the handlingof multi-dimensional and heterogeneous information in dynamic settings can
be performed.
 Easily identifies trend and patterns
 Fast processing and prediction inreal time
 Tasks are implementedautomatically easily
 Makes better decision
Phosphorous It promotes the root growth of plants and make them
to withstand low temperature.
Potassium It is a macronutrient present in the soil matter.
Organic
Carbon
It has an importantrole in the physical, chemicaland
biological function of agricultural soils.
Ferrous Soil also consists of major iron content. Iron richsoil is
acidic in nature.
Zinc It helps the plantsto produce chlorophyll.
Boron The reproductive and vegetative growth of plants isaffected
by the deficiency of boron.
Biological Micro- organism
biomass of C
and N
Microbes decompose soil organic matter releasing carbon
dioxide and plant available nutrients.
NaturalManure Manure like vegetable waste and the animal excreta that
saturates the organic matter insoil for plant growth.

The contribution of this paper is constructed as follows:
Section 2 elaborates literature survey for research. Section 3 depicts themethodology and section 4 discusses the result of
the proposed model and section 5 concludes the research paper by summarizing our work.
2 Related Work
This step involves breaking down the system into separate components to assess the situation, analysing the project's
goals, breaking down what needs to be established, and attempting to engage clients to identify particular requirements
[6].
Based on the nutrient present in the soil, the soil fertility would be predicted [7].
For crop analysis, we need to monitor various physical parameters such as Texture of soil, pH level or Electrical
Conductivity etc.
Overall analysis of soil is being carried outin the project based on the respective parameters. Crop selection method has
been developed for season wise cropprediction [8]. Therefore, based on Kharif (crops which are sown at the beginning of
the rainy season, e.g., between April and May.), Rabi (crops that are sown at the endof monsoon or at the beginning of
winter season, e.g., between September and October. These crops are known as monsoon crops.) and Zaid (short season
between Kharif and Rabi season in the months of March to July) the seasonal crops will be predicted [9]. For prediction,
they have compared and analysed differentalgorithms. One of the main factors that affect crop growth is texture of the soil.
Future vision & Scope
In the future we are expecting to work with the real time data/primary data. Currently we operate our research on
secondary data due to some limitations and lack of resources. We are planning to deploy our algorithm to the cloud so that
any device without prior training and testing of thedata can use it. We can introduce new machine learning algorithms and
tactics to improve the accuracy of the testing Dataset As currently we have a stable maximum accuracy. We can wide
spread our model suitable to various climatic conditions and zones so that it does not stick to a particular zone. We have
planned to provide a hard copy of report or the hard copy of compatible crop to the farmers andto the users. This saves
their time and money too.
3 Methodology
Sample of soil from SHC has been takenfrom secondary source to train the model. After, we removed the unwanted
attributesthat don’t contribute in the training model. Pre-process the data by adding newattributes to
help in further processing.“Soil type” is added on the basis of pHvalue of the soil. “Soil fertility” is addedon the
basis of B, C ,N content. Afterward,we divided the data into 80-20 rule fortraining and testing of the data. In the nextstep
we have used three ML models:Decision Tree Classifier, Random Forest,KNN to compare their respective accuracy.
The model with better accuracy than othertwo’s has been accepted for training.
Fig 3.1 shows the step by step flow chartof methodology accepted by our model.

Figure 3.1: Flowchart of the Methodology
Figure 3.2: Input Data Column
Figure 3.3: Data after Processing

Figure 3.4: Data ML Model
4 Results:
Gather the data recorded in Krishi Vigyan Kendra for the soil health report of thearea[10]. Check the attributes suitable for
the processing of the soil testing. On the basis of selected attributes removesunwanted attributes and add some attributes
like (Soil type, soil fertility) to the existing data. Remove the outliers using box plot method. Divide the data into 80-20 to
train and test the ML model respectively.
Starting with the KNN model, train ourdataset on this model and we get the accuracy of the model. We have accuracy of
64.08% (where n=6). For a better accuracy we train our model under Decision tree classifier model, and get the accuracy
of 84.21%. To get a betteraccuracy of the model we train the model with Random Forest algorithm and get the accuracy of
93.6%.
The accuracy of the model signifies that for a set of well define attributes our modelcan predict a crop that fits 93.6% to
the fertility of the soil.
4.1 Decision Tree Classifier:
Decision Tree is a type of supervised learning used for classification and regression problems. It aims to build a model that
predicts the value of a target variable by learning simple decision rulesinferred from data nodes.
Figure 1: Working of Decision TreeClassifier
4.2 KNN CLASSIFIER
KNN stands for K-Nearest Neighbours andit is based on Supervised Learning technique. It is one of the simplest algorithms
of Machine Learning. It works on the similarity measure between the input data and the available data and put the new
data into the category which is similar to the available category. This algorithm is mostly used for the Classification
problems but either be used for Regression or for Classification. considered were KNN Classifier, Decision Tree, K nearest
neighbour and Random Forest and among all four algorithms, the accuracy rate for Random Forest was high.Accuracy for
each algorithm is shown in Table 2 below.

Table 2 Accuracy of different algorithms
Figure 2: Working of KNN Classifier
4.3 RANDOM FOREST
We compared our data model accuracy Random Forest is a popular machinelearning algorithm that belongs to the
supervised learning technique. Both Classification and Regression problems can be solved by Random Forest. It is a
process of combining multiple classifiersto solve a common problem and to improve the performance of the model
which is the concept of ensemble learning.
Figure 3: Working of Random Forest
Different algorithms are to be compared. Different algorithms gave distinct results on the same dataset. The algorithm
with the pre-published Research paper onsoil testing and prediction [11].
Table 3 Comparison between Methods
Algorithms Published Method Our model Method
Decision Tree Classifier 61.5% 84.21%
Random Forest 72.74% 93.6%
Algorithms Accuracy
K-Nearest Neighbour 62.46%
KNN Classifier 64.08%
Decision Tree 84.21%
Random Forest 93.6%

Graph is plotted by using the above result.
Figure 4: Working of Graph Comparison
Figure 5: Graph Comparison
This figure shows the accuracy comparison of our model that is named as new_readings and we compared thosevalues by
another research paper named asold_readings in our paper that is already published.
5 Conclusion
In this project analysis of soil based onmajor and minor nutrients present in the soil has been proposed using Machine
Learning Techniques. The project has highefficiency and accuracy in fetching the realtime dataset of soil components. The
project will assist the farmers in increasingthe agriculture yield and take efficient careof crop production as the stick will
always provide helping hand to farmers for gettingaccurate live feed of soil fertility and the type of the soil upto93.6%
accurate results. The project proposes a wise agricultural model in integration with Machine Learning. Machine
Learning have always mattered in agriculture domain. It is really challenging task because of highly localized nature of
agriculture information specifically distinct conditions. We have used number of algorithms such as Decision Tree
Classifier, KNN Classifier and Random Forest. After testing each algorithms we get accurate result with 93.6% by Random
Forest. The complete real-time and historical environment information is expected to help for betterment.

References
1. T. Venkat Narayana Rao, “Prediction of Soil Quality usingMachine Learning Techniques”, International Journal of
Scientific& Technology Research Volume 8 (2019), ISSN 2277-8616
2. P. Shubham, R. Prem, S. Swami, P. Sandip, ”Soil Analysis and Crop Prediction”, International Journal of Scientific
Research in Scienceand Technology (2020), ISSN: 2396-6011, Vol. 7, Issue 4, Pagenumber: 117-123
3. K. Vishal, K. Raushan, K. Shubham, Ajinkya, Prof. P. P. Jorvekar, “Agriculture soil Analysis for Suitable Crop
prediction”.
4. Kazheen I T, Adnan M A, Dilovan A Z (2021), ”Data Mining Classification Algorithms forAnalyzing Soil Data”, Asian
Journal of Research in ComputerScience, 8(2): 17-28
5. R. Jayalakshmi, M. Savitha Devi, “Predictive Model Construction for Prediction of Soil Fertility using Decision Tree
Machine Learning Algorithm”.
6. P. Vinciya, Dr. A. Valarmathi, “Agriculture Analysis for Next Generation High Tech Farming in Data Mining”,
IJARCSSE, vol. 6, Issue 5, 2016.
7. Sabri Arik, Tingwen Huang, Weng Kin Lai, Qingshan Liu, “Soil Property Prediction”, An ExtremeLearning Machine
Approach‖ Springer, vol. 3, Issue 4,666-680, 2015.
8. Sadia Afrin, Abu TalhaKhan, “Analysis of Soil Properties and Climatic Data to Predict Crop Yields ” (2018 IEEE ICIS
2018).
9. Singh “Crop Selection Method to Maximize Crop Yield Rate using Machine Learning Technique”2015 IEEE,
International Conference on Smart Technologies andManagement for Computing,Communication, Controls, Energy
and Materials (ICSTM).
10. S.K.D Krishi Vigyan Kendra, Baghra, Muzaffarnagar, established by Shri Shri 1008 Veetrag swami Kalyan Dev ji
Maharaj, December (1995)
11. G. Jay, I. Anurag, G. Shailesh, A. Vahida, “Soil Data Analysis Using Classification Techniques and Soil Attribute
Prediction”,“InternationalJournal of Computer Science”, Issue 9, June (2012)
12. Abishek.B, AkashEswar“Prediction of Effective Rainfall and Crop Water Needs using Data Mining
Techniques” (2017 IEEE International Conference on Technological Innovations in ICT For Agriculture and Rural
Development (TIAR 2017)).
13. Rajkomar, A., Lingam, S., Taylor, A. G., Blum, M., & Mongan, J. (2017). Highthroughput classification of radiographs
using deepconvolutional neural networks. Journal of digital imaging, 30(1), 95-101.
14. Tensorflow. (2019, July 10). Tensorflow/tensorflow. Retrieved August 04, 2019, from
https://github.com/tensorflow/tensorflow/tree/master/tensorflow/exam ples/label _image
15. Dyrmann, M., Karstoft, H., & Midtiby, H. S. (2016). Plant species classification using deep convolutional neural
network. Biosystems Engineering, 151, 72- 80.

Digital Soil Mapping using Machine Learning

Recommended

More Related Content

Similar to Digital Soil Mapping using Machine Learning (20)

More from IRJET Journal (20)

Recently uploaded (20)

Digital Soil Mapping using Machine Learning