11 VIII Aug 2023
https://doi.org/10.22214/ijraset.2023.55225
International Journal for Research in Applied Science & Engineering Technology (IJRASET)
ISSN: 2321-9653; IC Value: 45.98; SJ Impact Factor: 7.538
Volume 11 Issue VIII Aug 2023- Available at www.ijraset.com
Soil Fertility Prediction
Aishwarya Patil1, Varshini Kulkarni2, Sachin Desai3
Department of Master of Computer Applications, VTU/KLS Gogte Institute of Technology, Belagavi
Abstract: Soil fertility plays a important role in figuring out agricultural productiveness and sustainability. Traditional methods
of assessing soil fertility involve time-consuming and high-priced laboratory tests, restricting their scalability and real-time
applicability. To overcome these challenges, this study proposes a data-driven method utilizing machine learning techniques for
accurate and efficient soil fertility prediction. Several machine learning algorithms, inclusive of decision trees, Random Forests,
k nearest neighbors, and Gradient Boosting Machines (GBM), are employed to model the complex relationships among soil
properties and fertility. Feature selection techniques are carried out to identify the most influential soil parameters for enhanced
prediction accuracy and reduced model complexity.The outcomes demonstrate that machine learning models can appropriately
predict soil fertility, outperforming traditional approaches in terms of speed and cost-effectiveness. Moreover, the characteristic
selection process identifies key soil properties that have the most significant effect on fertility, offering valuable insights for
agricultural decision-making and targeted soil management. The proposed approach offers potential applications in precision
agriculture, enabling farmers to make knowledgeable choices regarding crop selection, nutrient management, and irrigation
strategies based totally on actual-time soil fertility predictions. By optimizing resource allocation and minimizing environmental
influences, this data-driven solution contributes to the promotion sustainable agricultural practices and guarantees food safety
for a growing global population. These are the essential nutrients that the crop requires for its growth pH nitrogen(N),
phosphorus(P), potassium(K), CaCo3, Organic Carbon, Organic matter, CEC (Cation exchange capacity) .
Keywords: Soil, Prediction, Fertile, Non Fertile
I. INTRODUCTION
Agriculture is one of the predominant career in India. huge populace of India relies upon agriculture as their most important source
of profits. With time, the demand for manufacturing has been multiplied exponentially[2].
one of the enormous elements that has an instantaneous impact on crop production and first-class is soil fertility and its nutrient
control. offering vegetation with the suitable amount
of nutrients on the right time is the key to a success crop production.[1]
Any malfunction of the soil homes impacts now not simplest agriculture however also the water cycle of the earth and to an extent
paves way for herbal calamities. The parameters that have an effect on the fertility are many, that are predicted in laboratories using
traditional methods, and its want of the
hour to revolutionize those estimation strategies of soil fertility using automated strategies.[5]
To have an powerful production of the crop and adding fertilizers inside the proper ratio within the soil, it is important for the
farmers to realize the soil nutrient composition. consequently, strength of soil nutrient evaluation has grow to be the need of
this today’s world.The machine learning algorithm plays a high function and offers faster and correct consequences[6].
Machine learning algorithms like Random forest, k-nearest neighbour, gradient boosting were used for results in soil nutrient
analysis.
The motto is to expect the outcomes accurately thru the implementation of the ML approach.[6]
Fig 1. Methodology
©IJRASET: All Rights are Reserved | SJ Impact Factor 7.538 | ISRA Journal Impact Factor 7.894 | 1241
International Journal for Research in Applied Science & Engineering Technology (IJRASET)
ISSN: 2321-9653; IC Value: 45.98; SJ Impact Factor: 7.538
Volume 11 Issue VIII Aug 2023- Available at www.ijraset.com
II. PROPOSED SYSTEM
The proposed system predicts the amount of soil nutrients specifically N,P,K the use of the trained data set which
has various parameters and the more than one linear regression set of rules.for this reason, understanding those values of
soil nutrients, it might be easier for the farmers for adding fertilizer in a proper variety and bring higher yield.[6]
1) Methodology: Soil fertility is measured in terms of presence/ absence of nutrients present in the soil. chemical properties plays
a significant role in deciding the fertility status.The data set is being imported, it consists of parameters such as pH nitrogen(N),
phosphorus(P), potassium(K), CaCo3, Organic Carbon, Organic matter, CEC (Cation exchange capacity)those parameters play
a massive position in development and increase of a plant. thus, these parameters are used as variables where the nitrogen,
potassium and phosphorous bureaucracy the goal variable respectively.[6] The data set is skilled the use of the
MLR model.This version is used due to its performance to handle a couple of variables. The scatter plot and pairs characteristic
that is used for the advent of graph of the model that shows the contribution of each parameter. There
are extraordinary values of each nitrogen, potassium and phosphorous values predicted after giving the parameters pH,
moisture, electric conductivity, temperature as an input. The expected values are then used for growing graphs for N, P, K
respectively.[6]
2) Data Pre-processing: The data set acquired was in row form and consisted of redundant data. some data values were missing.
For correct prediction, the rows containing lacking values and the outliers within the data-set have been removed[2
3) Python Implementation: After pre-processing of the information, the machine learning algorithms- K nearest neighbour,
Random forest and gradient boosting had been implemented on the data set the usage of Python. The classes- good, average
and poor have been transformed to numerical form for the ease of prediction. in the k nearest neighbour algorithm prediction
has been done.[2]
4) Accuracy: The data set become divided into 5 folds and the average accuracy was calculated for each fold. The
final accuracy turned into obtained by way of calculating the Average accuracy of the five folds.[2]
5) Data Visualization: Data visualization changed was done using the Python library Matplotlib. The variables were additionally
converted to numerical form as in line with the requirement.
Fig 2data flow diagram
III. ALGORITHM
A. Random Forest(RM)
Random forest is a popular algorithm for soil fertility prediction using machine learning. it is a supervised learnin
g algorithm that may be used for both type and regression tasks.each decision tree inside the Random forest is grown with the aid of
recursively partitioning the data based on different features and splitting criteria. The splits are decided via maximizing the records
advantage, Gini impurity, or every other appropriate metric. The tree continues to develop until a distinct termination situation is
met.
©IJRASET: All Rights are Reserved | SJ Impact Factor 7.538 | ISRA Journal Impact Factor 7.894 | 1242
International Journal for Research in Applied Science & Engineering Technology (IJRASET)
ISSN: 2321-9653; IC Value: 45.98; SJ Impact Factor: 7.538
Volume 11 Issue VIII Aug 2023- Available at www.ijraset.com
Fluorescence can be utilized as a valuable tool for assessing soil fertility and health. Specifically, fluorescence spectroscopy is an
analytical technique that measures the fluorescence emitted by organic matter present in soil samples when they are exposed to
ultraviolet or visible light. the study's finding that induced fluorescence can be used to predict the nitrogen rate directly with an
overall accuracy of 0.78 (or 78%) is highly significant. The ability to predict the nitrogen rate directly from soil samples using
fluorescence measurements can have practical implications for farmers and agriculture in general[7].
Fig 2: output of random forest
In our test data there are 11 data are present, after applying random forest algorithm, we found 5 are Fertile, 2 are Non Fertile, 5 are
Medium Fertile.
B. K Nearest Neighbor
For each instance within the training records, the algorithm calculates the distance between the input feature of the instance and the
input features of the data point for which a prediction is required. The most normally used distance metric is Euclidean distance.
The algorithm selects the k nearest neighbors from the training data based on the calculated distances. The value of k nearest
neighbors from the training data based on the calculated distances . It determines the variety of neighbors so that it will make a
contribution to the prediction.
Fig 3. output of k-nearest neighbour
©IJRASET: All Rights are Reserved | SJ Impact Factor 7.538 | ISRA Journal Impact Factor 7.894 | 1243
International Journal for Research in Applied Science & Engineering Technology (IJRASET)
ISSN: 2321-9653; IC Value: 45.98; SJ Impact Factor: 7.538
Volume 11 Issue VIII Aug 2023- Available at www.ijraset.com
C. Gradient Boosting
Gradient Boosting is an ensemble device studying approach used for both regression and classification tasks. It builds
a strong predictive version with the aid of combining the predictions of more than one vulnerable newbies (usually decision trees) in
a sequential way. The idea behind gradient boosting is to accurate the mistakes made with the aid of the preceding model new
release via fitting next fashions to the residuals of the preceding ones.
The predictions from the new model are brought to the predictions from the previous version, updating the overall predictions of the
ensemble.
Fig 4: output of gradient boosting
Fig 5:confusion matrix
IV. BENEFITS OF SOIL FERTILITY PREDICTION
A. Crop Yield Optimization
By appropriately predicting soil fertility, farmers can optimize their crop yield capability. Soil fertility prediction fashions can help
decide the top of the line quantity and sort of fertilizers wanted for particular crops, making sure that the soil has the important
nutrients for surest growth and productiveness. This records enables farmers to make informed choices concerning fertilization
practices, leading to accelerated crop yields and advanced aid control.
B. Nutrient Management
Soil fertility prediction can resource in green nutrient management. via predicting the nutrient content and availability inside the soil,
farmers can tailor their fertilizer programs to healthy the particular nutrient necessities of vegetation. This reduces the chance of
over-fertilization, that can damage the surroundings, whilst making sure that the vegetation get hold of k nutrition for wholesome
growth.
C. Environmental Impact Assessment
Soil fertility prediction models can assist check the potential environmental affects of agricultural practices. through thinking about
factors inclusive of nutrient runoff and leaching, those models can estimate the risk of nutrient pollutants in nearby water bodies.
This records lets in policymakers and land managers to broaden suitable mitigation techniques to minimize environmental harm.
©IJRASET: All Rights are Reserved | SJ Impact Factor 7.538 | ISRA Journal Impact Factor 7.894 | 1244
International Journal for Research in Applied Science & Engineering Technology (IJRASET)
ISSN: 2321-9653; IC Value: 45.98; SJ Impact Factor: 7.538
Volume 11 Issue VIII Aug 2023- Available at www.ijraset.com
D. Precision Agriculture
Soil fertility prediction may be integrated with different technologies, including far off sensing and geographic records syst
ems (GIS), to enforce precision agriculture practices. through mapping soil fertility versions across a area, farmers can follow site-
precise fertilizers, making sure that vitamins are dispensed according to the precise needs of every soil vicinity. This targeted ap
proach can optimize useful resource usage, reduce fertilizer wastage, and enhance standard crop health and productivity .
E. Land Use Planning and Decision Making
Soil fertility prediction models play a essential role in land use making plans and selection making. through assessing soil
fertility characteristics, land managers could make informed selections approximately land suitability for numerous functions, along
with agriculture, forestry, or city improvement. Soil fertility prediction allows identify regions with low fertility that may require
soil amendments or alternative land uses, stopping capability crop screw ups and maximizing land productiveness.
F. Soil Conservation
Predicting soil fertility is critical for soil conservation efforts. with the aid of figuring out areas with low fertility, erosion-
inclined soils, or nutrient imbalances, land managers can implement suitable soil conservation measures. these may
additionally consist of terracing, contour plowing, cover cropping, or targeted soil amendments to enhance soil structure, lessen ero
sion, and décorate nutrient content.
V. CHALLENGES IN SOIL FERTILITY PREDICTION
1) Complexity of Soil Systems: Soil fertility is stimulated by using various factors inclusive of organic depend content material,
nutrient levels, soil structure, pH, microbial activity, and greater. The interactions between these elements Make it a
complicated system to model appropriately.
2) Spatial Variability: Soil fertility can vary significantly over small distances, leading to spatial heterogeneity. Cor
rect prediction requires information at a fine spatial decision, which can be costly and time-consumin g to accumulate.
3) Limited Data Availability: Soil data collection is an expensive and time-eating procedure. in lots of regions, soil records may
be scarce or old, which makes building correct predictive models difficult.
4) Data Quality and Consistency: Soil records is often accumulated the usage of diverse strategies and technology, main to
inconsistencies in facts exceptional, codes, and scales. Integrating and standardizing statistics from specific assets can
introduce noise and errors in the predictions.
5) Non-Linear Relationships: The relationships among soil properties and soil fertility signs are frequently non-linear. Traditi
onal linear models might not capt ure these complicated relationships correctly.
6) Over-fitting and Generalization: Soil fertility prediction models can be prone to over-fitting, mainly while dealing
with constrained data or using complicated algorithms. Ensureng the version generalizes properly to new, unseen records is I
mportant for realistic packages.
Fig 6: error rate
©IJRASET: All Rights are Reserved | SJ Impact Factor 7.538 | ISRA Journal Impact Factor 7.894 | 1245
International Journal for Research in Applied Science & Engineering Technology (IJRASET)
ISSN: 2321-9653; IC Value: 45.98; SJ Impact Factor: 7.538
Volume 11 Issue VIII Aug 2023- Available at www.ijraset.com
7) Temporal Variability: Soil fertility can change over time because of various natural and anthropological factors, which
include weather changes, land use, and agricultural practices. Modeling these dynamic changes accurately is challenging.
8) Handling Missing Data: Soil datasets may have missing values, which need to be treated accurately. Managing missing data w
ith out introducing biases is a critical aspect of constructing s trong predictive models.
Fig.7 Block Diagram of the Proposed System
VI. FUTURISTIC SOLUTION
Information about pesticides may be Introduced to the crop to improve its growth and to prevent the flowers from being affected
by the pests. A graph may be generated that gives the facts about The crops which can be presently planted via the other farmers in a
particular area. the use of those facts farmers can get a clean concept about the crop to be planted.[4]
This could further be extended by means of developing a mobile application where the values of N, P, k,
CaCo3, organic Carbon, natural matter, CEC that is acquired from the model could be displayed inside the application developed.
The values may be stored in a database and displayed in the app. consequently the system
advantages the farmers and helps them taking right choices for growing plants.
Fig 8: Heatmap
©IJRASET: All Rights are Reserved | SJ Impact Factor 7.538 | ISRA Journal Impact Factor 7.894 | 1246
International Journal for Research in Applied Science & Engineering Technology (IJRASET)
ISSN: 2321-9653; IC Value: 45.98; SJ Impact Factor: 7.538
Volume 11 Issue VIII Aug 2023- Available at www.ijraset.com
VII. CONCLUSION
soil fertility prediction the usage of machine learning is a promising field that has the potential to improve crop productiveness and
yield. through the usage of various machine learning techniques inclusive of k-nearest neighbors, decision trees, Random forest and
Gradient boosting, we can analyze soil data which include nutrient content material, pH nitrogen(N), phosphorus(P) and the
potassium(k),CaCo3, organic Carbon, organic matter, CEC content present in the soil to appropriately predict soil fertility stag
es. This may help farmers make informed selections about fertilization and crop management practices, resulting
in more efficient and sustainable agriculture.Overall, machine learning can be a valuable tool in improving soil fertility management
and contributing to sustainable agriculture.
REFERENCES
[1] Jeremy Aliparo, “IOT-BASED ASSESSMENT AND MONITORING OF NPK CONTENT AND FERTILITY CONDITION OF SOIL”, in 2022 IEEE regiin
10 conference, Available: https://ieeexplore.ieee.org/search/searchresult.jsp?newsearch=true&queryText=soil%20fertility%20prediction
[2] Pranay Malik, “COMPARATIVE ANALYSIS OF SOIL PROPERTIES TO PREDICT FERTILITY AND CROP YIELD USING MACHINE LEARNING
ALGORITHMS”, 2021 11th International Conference on Cloud Computing, Data Science & Engineering (Confluence) | 978-1-6654-1451-7/20/$31.00 ©2021
IEEE DOI: 10.1109/Confluence51648.2021.9377147, Available:https://ieeexplore.ieee.org/search/searchresult.jsp?newsearch=true&query Text=soil%
20fertility%20prediction
[3] G Sai Pravallika, “PROFICIENT SMART SOIL BASED IOT SYSTEM FOR CROP PREDICTION”, Proceedings of the Second International Conference on
Inventive Research in Computing Applications (ICIRCA-2020) Available:https://ieeexplore.ieee.org/search/searchresult.jsp?newsear ch=true&query
Text=soil%20fertility%20predictio
[4] P. S. Vijayabaskar, Sreemathi.R, Keertanaa.E, “CROP PREDICTION USING PREDICTIVE ANALYTICS”, 2017 International Conference on Computation
of Power, Energy, Information and Communication (ICCPEIC) Available:https://ieeexplore.ieee.org/search/sear chresult.jsp?newsearch=true&q ueryTex
t=soil%20fertility%20prediction
[5] Sareena Rose , S. Nickolas , S. Sangeetha, “MACHINE LEARNING AND STATISTICAL APPROACHES USED IN ESTIMATING PARAMETERS THAT
AFFECT THE SOIL FERTILITY STATUS: A SURVEY”, Second International Conference on Green Computing and Internet of Things (ICGCIoT)
Available:https://ieeexplore.ieee.org/search/searchresult.jsp?newsearch=true&queryText=soil%20fertility%20prediction
[6] Madhumathi R ,Arumuganathan T , Sneha Iyer R , Shruthi R , Shruthhi K,“SOIL NUTRIENT ANALYSIS USING MACHINE LEARNING TECHNIQUES”,
In National E-Conference on "Communication, Computation, Control and Automation" ( CCCA-2020) Available:https://ieeexplore.ieee.org/search/searchresul
t.jsp?newsearch=true&queryText=soil%20fertility%20predictio
[7] Louis Longchamps , Dipankar Mandal, “ASSESSMENT OF SOIL FERTILITY USING INDUCED FLUORESCENCE AND MACHINE LEARNING” ,2022
by the authors. Licensee MDPI, Basel, Switzerland.
©IJRASET: All Rights are Reserved | SJ Impact Factor 7.538 | ISRA Journal Impact Factor 7.894 | 1247
