See discussions, stats, and author profiles for this publication at: https://www.researchgate.

net/publication/346627389

Crop Recommendation System

Conference Paper · October 2020

CITATIONS READS

0 1,249

1 author:

Thewahettige Harinditha Ruchirawya

Sri Lanka Institute of Information Technology
1 PUBLICATION   0 CITATIONS   

SEE PROFILE

All content following this page was uploaded by Thewahettige Harinditha Ruchirawya on 04 December 2020.

The user has requested enhancement of the downloaded file.

 International Journal of Computer Applications (0975 – 8887)
Volume 175– No. 22, October 2020

Crop Recommendation System

Pradeepa Bandara Thilini Weerasooriya Ruchirawya T.H.

Lecturer Assistant Lecturer Undergraduate
Sri Lanka Institute of Information Sri Lanka Institute of Information Sri Lanka Institute of Information
Technology Technology Technology
Sri Lanka Sri Lanka Sri Lanka

W.J.M. Nanayakkara Dimantha M.A.C Pabasara M.G.P

Undergraduate Undergraduate Undergraduate
Sri Lanka Institute of Information Sri Lanka Institute of Information Sri Lanka Institute of Information
Technology Technology Technology
Sri Lanka Sri Lanka Sri Lanka

ABSTRACT agriculture enable businesses and farmers to be additional

Automating agricultural aspects is a mechanical process with profitable, efficient, safer, and more environmentally friendly.
or without human intervention in agriculture. Due to less The rise of digital agriculture and its connected technologies
space of domestic lands, it has become an important area of has opened a wealth of latest knowledge opportunities.
choosing the most suitable crops based on prevailing factors Remote sensors, cameras, and alternative connected devices
in the selected area. In Sri Lankan even though there are will gather data twenty-four hours per day over a complete
enough knowledge, techniques, and methods which are done farm or land. These will monitor plant health, soil condition,
manually available in agriculture, there is not any system in temperature, humidity, etc. the quantity of information these
which the environmental factors are detected and suggests the sensors will generate is overwhelming. this enables farmers to
user which crop type is best for farming. This paper is achieve a far better an improved understanding of state of
consisting of a theoretical and conceptual platform of matters on the bottom through advanced technology which
Recommendation system through integrated models of will inform them additional regarding their situation more
collecting environmental factors using Arduino accurately and quickly.
microcontrollers, Machine learning techniques such as Naïve
Bayes (Multinomial) and Support Vector Machine (SVM), The environmental data that is gathered by remote sensors are
Unsupervised machine learning algorithm such as K-Means processed by algorithms and statistical data which will be
Clustering and also Natural Language Processing (Sentiment understood and helpful to farmers for decision makings and
Analysis) concerned with the Artificial Intelligence to keep track of their farms. The more inputs and statistical data
recommend a crop for the selected land with site-specific collected, and higher the algorithmic rule is at predicting the
parameters with high accuracy and efficiency. It has been a outcomes. And the aim is that farmers will use these
major problem to identify what to grow, any man has technologies to attain their goal of improved harvest by
adequate space in the owner’s land. Not only domestic lands creating better selections within the field. By implementing
but also for farming lands. Why it has become a problem is the system of temperature, soil hydrogen ion concentration
that environmental factors such as temperature, water levels, and soil wetness detection, the information captured are
and soil conditions are uncertain as they change from time to processed with an explicit algorithmic rule and passed to a
time. Due to these problems, this solution of crop centralized database that is connected to different modules of
recommendation system predicts the user, what crop type the research, so the main system will predict the most
would be the most suitable for the selected area by collecting effective crop kind that the farmer should grow to require the
the environmental factors for plant growth and processing most outcome of the crop kind that is farmed in a home
them with the trained sub-models of the main of the system. garden or the respectable land area.

General Terms 2. LITERATURE SURVEY

IOT – Internet of Things In this literature review of this project, the team sought out
and studied various patents, research papers, documents, and
Keywords newspapers and magazine articles from various scenes. The
Agriculture, Machine Learning, Arduino, Natural Language paper [1] states requirements and why they tend to move into
processing, Farming precision agriculture [2] which is due to globalization are
discussed. Precision agriculture is site-specific farming.
1. INTRODUCTION Though Precision agriculture has shown an improvement with
Modern farms and agricultural operations ar taken place more time, there exist some issues. As mentioned above site-
totally different than those many decades agone, primarily due specific methods of such systems are needed to be supervised
to advancements in technology, as well as sensors, devices, to get an improved result. Only a few of the outcomes are
machines, and knowledge technology. Today’s agriculture provided a particular result. Nevertheless, the situation is
habitually uses refined technologies like robots, temperature farming is indispensable since if any default or a mistake
and wetness sensors, aerial pictures, and GPS technology, and occurs, it might lead to serious damage to resources and as
lots of complex IOT devices. These advanced devices in well as the plants.

22
 International Journal of Computer Applications (0975 – 8887)
Volume 175– No. 22, October 2020

In this research, it is proposing a system where the major Table 1. Sample data set
factors are taken into consideration at the same time and come
up with a solution so that the system will not be complicated Crop Temp. Hum. pH Sun. Moist.
for the user. As mentioned above in the sentence, the major Bean 26 63 7 8 63
factors taken into consideration at once is unlike other models
proposed in previous researchers, this system considers all the Lettuce 23.8 60 6.5 8 60
major factors which are essential for plant growth, are
processed together using various algorithms whereas the other Carrot 26 98 6.5 8 98
models consider only parameters at once keeping the other Cabbage 29.4 95 7.1 5 95
factors constant.
Beet 30 99 6.5 6.5 6.5
As an example, some tests are carried out to find the rate
evaporation and how the plant growth is affected when there’s
no sufficient water present. And a derived equation is
presented as a result [3]
ETo = K pan × E pan
ETo: reference crop evapotranspiration
Kpan: pan coefficient E pan: pan evaporation
Even though an equation is proposed, there are some
limitations [4]. Mainly this could be done only for a land with
a less area. This is not suitable for commercialization as the
profit will be low when you use less area for cultivation. The
second limitation is the average rainfall of Sri Lanka mostly
suitable for many crop types to grow without any deficiency.
Therefore, the water level alone itself is not a big issue
whereas other factors are. If there’s enough water for plants
but no temperature that would be a problem as the major
environmental factors have a mutual relationship with each Fig 1: Data plot of soil moisture using Linear Algorithm
other in plant growth.
The ultimate goal of the previous researches is also to predict
the best crop type. But once the farmer or user has cultivated
the predicted crop type, the system’s job is done. But in the
system that proposed from this paper, has a feedback system
as well. Even after suggesting the best crop type, the system
can track the plant growth and it provides feedback if the farm
is malnourished. So that the user can take necessary
precautions prior.

3. METHODOLOGY
3.1 Dataset Collection
It is needed to have a certain temperature, humidity, soil pH,
sunlight, and soil moisture for a plant to be grown healthy. To
receive to good harvest those conditions should be satisfied.
But those conditions may vary according to the plant varieties. Fig 2: Data plot of Sunlight (lux) using Linear Algorithm
The initial data set is collected from the Department of
Agriculture Sri Lanka [5], other agriculture books, 3.2 Collecting Environment Factors
Agricultural web sites [6], and other reports and research To compare and predict the initial data set, the environmental
papers. This initial data set was used to train the crop factors needed to be gathered. To collect environmental
recommendation model so that the accuracy would be factors, Arduino microcontrollers [7] are used. Since both the
increased. temperature and humidity sensors are comprised of a single
micro-controller, four sensors are used to collect data. They
are sunlight intensity sensor, soil moisture sensor, soil pH
sensor, and humidity and temperature sensor. Those sensors
are connected with an Arduino Wi-Fi module and gathered
data is sent to the database. The gathered data are cleaned and
processed by using clustering and other algorithms to pass the
values to the next component of crop recommending and
stored in the database.

23
 International Journal of Computer Applications (0975 – 8887)
Volume 175– No. 22, October 2020

Fig 3: Arduino Micro-Controllers Connectivity

3.3 Crop Prediction soil condition or any other changes in the selected land. But to
Since the environmental conditions differ from region to avoid these factors affecting the crop prediction, the farmer’s
region, a machine learning model is used to predict the best feedback system is included in the system.
crop type for the selected land. To train the crop
recommending model with the data collected from the Once the crop type is recommended, the farmer is asked for
Arduino sensors, machine learning algorithms [8] are used to details and feedbacks regularly through the mobile application
identify the best crop to cultivate with the highest probability to guide the farmer with necessary precautions. To provide the
of growing. Naïve Bayes & Support vector machine necessary feedback, the feedback system is used in the mobile
algorithms are used to select the best crop type. application by selecting the crop type. By this, the overall
accuracy and the reliability of the product are increased with
From this model, it decided what type of crops that the farmer time.
should grow. This is done by analyzing factors of humidity,
temperature, soil moisture, pH level, and sunlight. Mainly the 4. RESULT
system suggests 4 crop types by analyzing the above- The proposed system was successfully implemented in a
mentioned factors using two machine learning algorithms. selected land. The data sampled from the sensors at an
interval of one hour on a normal sunny day is presented in
Naïve Bayes [9] - Naive Bayes is a technique for constructing Table I. Based on the tests conducted and the data collected, it
classifier models that assign class labels to problem instances can suggest the best crop to cultivate for maximum harvest.
which are represented as vectors of feature values, where the Using the feedbacks collected from the farmer the accuracy of
class labels are drawn from some finite set. the predictions is sharpened by neglecting the invalid data. As
Support Vector Machine (SVM) [10] - The objective of the an example, if the farmers are continuously providing the
support vector machine algorithm is to find a hyperplane in N- negative feedback for cultivating strawberry in Galle, the
dimensional space (N — the number of features) that system itself learned, and the accuracy of the final output is
distinctly classify the data points. increased which means in future it will not suggest strawberry
to grow in Galle. The farmer can enter his own feedback using
3.4 Monitoring and Feedback his native language. Since the most Sri Lankans use Sinhala
This proposed system product would mainly identify four and English, the custom libraries are implemented, by which
types of crops according to the environmental factors of the the system identifies the language significantly by using
selected plot of land. But the reason to obtain a probability of natural language processing algorithms and the feedback is
more than 90% for the above-identified crops would be the identified.

24
 International Journal of Computer Applications (0975 – 8887)
Volume 175– No. 22, October 2020

According to the prevailing environmental factors in the agriculture consultancy measures, this system will definitely
selected land, the best suitable four crop types are suggested be a torchbearer.
to the farmer.
6. FUTURE WORKS
It can do a lot of additional features to the system. Currently,
it takes necessary environmental factors as inputs and
suggests a very suitable crop to be cultivated. But as the next
level, the Automation part can be added as the response
system to the feedback. This can be modified to control the
humidity, water level, etc. according to the need of the farmer.
Presently it takes all environmental factors as inputs to the
system, but as an additional feature, an algorithm can be
Fig 4: A sample output given to the farmer implemented to predict the one factor using another two
factors. (Example – predicting the soil pH level from soil
Figure 4 shows a sample result where the farmer gets, once moisture and sunlight), so that the initial cost of setting up the
the environmental factors are entered. PLU code is an sensors would be less and can be easily maintained.
attribute that is used to identify the crop uniquely.
7. REFERENCES
The overall accuracy of the proposed system is more than [1] Lakshmi.N, Priya.M, Sahana Shetty, and Manjunath C.
92%. As the farmer continue to use this, the more data system R, Crop Recommendation System for Precision
feeds, the more accurate it will be. Therefore. by the time of Agriculture, vol. 6 Reading, IND: International Journal
long usage, the farmer can obtain more the 95% of accuracy for Research in Applied Science & Engineering
from the whole system. Technology, 2018. [Online] Available:
Table 2. Accuracy level of the models in the system www.ijraset.com.Ding, W. and Marchionini, G. 1997 A
Study on Video Browsing Strategies. Technical Report.
Name of the Module Accuracy University of Maryland at College Park.
Temperature and Humidity
92% [2] Remi Schmaltz, “What is Precision Agriculture”, April
detection model 2017. [Online]. Available:
2Sunlight intensity, soil pH and https://agfundernews.com/what-is-precision-
95%
soil humidity detection model agriculture.html [Accessed Feb.23, 2020].
Crop recommending system 90% [3] C. Brouwer and M. Heibloem, Irrigation Water
Monitoring and feedback system. 96% Management: Irrigation Water Needs, manual 6 Reading,
ITALY: Food and Agriculture Organization of the
United Nations, 1987. [Online] Available:
http://www.fao.org/3/s2022e/s2022e07.htm#TopOfPage.
5. CONCLUSION
In a modern environment with less space and less knowledge [4] A. de Carbon, "PRECISION AGRICULTURE: ITS
of agriculture, all the factors are considered from the BENEFITS AND LIMITATIONS", carrhure, 2019
perspective of farmer and plant, and the farmer is properly [Online]. Available: https://www.carrhure.com/precision-
guided until the harvesting. Before selecting any plant to grow agriculture-benefits-limitations/ [Accessed: Feb.25,
it is important to have the knowledge and an understanding of 2020].
the factors that affect the cultivation and how to maintain or
control them. From this system, these above-mentioned [5] Department of Agriculture Sri Lanka (2015), ‘Crop
factors are automatically processed and select the crop type to Suitability Recommendation for Grama Niladhari
be cultivated. Divisions in Sri Lanka’, Natural Resources Management
Center, Peradeniya.
Once the plant is cultivated, the farmer is asked for feedback
regularly with a time interval of one month. From this [6] Marie ionnotti (2020), ‘Outdoors & Gardning’, The
feedback taken, the system self-trained, and the accuracy is Spruce Web Site
improved with time and data collected. From this system, the [7] WolkWriter (2019), ‘Periodically measure environmental
guidance of a specialist is not needed, and the maintenance is conditions and send them to WolkAbout IoT Platform to
less. Thus, implementing this system will not have any monitor the environment remotely.’, ARDUINO
additional monetary impact on the user. PROJECT HUB
Since the average land space for a person in Sri Lanka is less, [8] Simon Tavasoli (202), ‘10 Machine Learning Algorithms
from the data gathered by sensors it is evident that this You Need to Know’, Simplilearn Solutions
proposed system has above 95% accuracy and suitable for
both rural and urban areas. With the time of the accuracy [9] Sunil Ray (2015), ‘6 Easy Steps to Learn Naive Bayes
would be more increased significantly as the significant Algorithm with codes in Python and R’, Analytics
amount of data is collected and the system self-trained by the Vidhya
time of approximately one year. The experiments conducted [10] Rohith Gandhi (2018), ‘Support Vector Machine —
over the 4-6 months proved the reliability and accuracy of the Introduction to Machine Learning Algorithms’, Toward
system. When the whole world is looking forward for new Data Science Web Site

IJCATM : www.ijcaonline.org
25

View publication stats