Machine Learning Prediction Analysis using IoT for Smart Farming

Abstract

Commonly, the implementation of agricultural practices (e.g., ploughing, sowing, watering, pests control, and harvesting.) purely depends on climate change, recommendations from previously experienced rules, and Governmental policies. For fulfilling Term smart farming, i), we employed real-time applications over sensors to capture climate changes of soil and atmosphere. ii) we defined agriculture practice rules by applying machine learning techniques over the last five years data iii) By federations of real-time data from the field sensors and rules, we define the time for implementation of the practice. This federation eliminates many malfunctions in old ways of smart farming for precision agriculture.

Full text

6482
ISSN 2347 – 3983
Volume 8. No. 9, September 2020
International Journal of Emerging Trends in Engineering Research
Available Online at http://www.warse.org/IJETER/static/pdf/file/ijeter250892020.pdf
https://doi.org/10.30534/ijeter/2020/250892020
Machine Learning Prediction Analysis using IoT for
Smart Farming
Abdul Rehman1, Jian Liu2, Li Keqiu3, Ahmed Mateen4* and Muhammad Qasim Yasin5
1,2,3College of Intelligence and Computing, Tianjin University, Tianjin Key Laboratory of Advance Networking, Tianjin, China.
*4Department of Computer Science, University of Agriculture Faisalabad, Pakistan, 38000
Email: ahmedbuttar@uaf.edu.pk
5College of Intelligence and Computing, Tianjin University, Tianjin Key Laboratory of Cognitive Computing and Application,
Tianjin, China
ABSTRACT
Commonly, the implementation of agricultural practices
(e.g., ploughing, sowing, watering, pests control, and
harvesting.) purely depends on climate change,
recommendations from previously experienced rules, and
Governmental policies. For fulfilling Term smart
farming, i), we employed real-time applications over
sensors to capture climate changes of soil and
atmosphere. ii) we defined agriculture practice rules by
applying machine learning techniques over the last five
years data iii) By federations of real-time data from the
field sensors and rules, we define the time for
implementation of the practice. This federation eliminates
many malfunctions in old ways of smart farming for
precision agriculture.
Keywords: Agriculture, IoT, Sensors, Smart Farming,
Raspberry Pi3.
1. INTRODUCTION.
Pakistan is an agro-based country, 47 % of the total land area
is for agriculture, and 61% of rural community make their
earnings directly or indirectly from agriculture [1]. It is
playing an essential contribution in the economy.
With the regular expansion of industrializations and
urbanizations, the agricultural area is continuously reducing.
It is boosting the demand for food due to the high growth rate.
They are still practicing traditional agriculture with the old
traditional methodology of cultivation. Which are needed to
be replaced with new emerging technologies, e.g., field
sensors implementations [2]. data analytics, experienced bases
decision systems. Agri-data of the last fifteen years about pest
attacks, agricultural practices regarding different zones is
available. Smart farming is the demand of the era to meet the
food demand by boosting crop production. The authors [3]
implemented big data analytics to formulate agriculture
data.The stored agri-data is needed to be cleaned and stored to
be used for valuable decision support systems. In this
manuscript, we import data in heterogeneous from
heterogeneous systems. For example, the old agri-data is
available in CSV, TXT, and relational database format.
Secondly, we collect real-time data with IoT sensors from
agriculture. This real-time data stream is composed of soil
temperature, atmospheric temperature, and humidity. The
practices plan is formulated with the information extracted
from the following sources e.g.
1) Real-time data streams from Sensors
2) Information about the sensor locality
3) Information extraction from five years old data
4) Climate forecast by the government
5) Watering System
The Agricultural Big source of heterogenous Data Because the
Agricultural Data is also being collected from different
sources like public data, private data, industrial data, and
Governmental data. So, data privacy and security are also
required for private data and industrial data [4]. For such [3] a
significant and diversified data, we need to develop a large
cloud environment for data management and analytics [5].
The apache-spark [6] based cloud environment is constructed
for data cleaning, collecting, storing, and data analytic &
Machin learning.
2. BACKGROUND.
2.1 Precision farming
Precision farming is the term that refers to the effective and
efficient use of limited inputs to get more outputs. It is a new
way to employ digital technology to optimize agriculture
practices. Different trends of technologies are changing the
parameters and shapes of precision agriculture. The major
trends are IoT, weather forecast, and big data technologies
The Internet of things has a significant impact in all fields; it
gives new automation [31]. It leads to new directions for
scientists to implement their research. Specifically, it
improves the efficiency of operational work. In agriculture, it
applications of agriculture [7]. For example, it made it
possible to attain real-time field data about soils and
atmospheric temperature. For large form, aerial imagery and
drones, and satellites are implemented as a tool of precision
agriculture. To make the precision agriculture more
sustainable, reliable, optimization and productive we
employed machine learning and data analytics to formulate
the quick decision.

Abdul Rehman et al., International Journal of Emerging Trends in Engineering Research, 8(9), September 2020, 6482 – 6487
6483
2.2 IoT in Agriculture
The main factors of sustainability and reliability can be
ensured with implementation of the smart farming concepts.
We can optimization agricultural practices by defining rules
by application of Machine learning and by recording the real
time rainfall, temperature, humidity and pest activity and
prediction. For predicting soil drying, soil conditions, soil
temperature, and soil moisture KNN ANN /MLP and RBF are
used [8]. The diseases identifications [9,28] is no more
unresolved. The agri-fields are monitored by satellites,
sensors and drones. The data is collected a for crop pest
management and irrigation system. The real time data we
need field sensors for soil temperature, humidity and
atmospheric temperature and humidity.
The remote sensing system is effective for detection of plant
disease and it is very cheap as compare to traditional methods
of pest scouting. In modern word, area which can be cultivated
is rapidly reducing. Sensor technologists and agriculture
scientists are working in collaboration to increase productivity
in agriculture field [10]. Crop yield predication techniques
have an important role in improving crop productivity because
they help the farmers in timely decisions about production,
storage and making of crops as well as risk management [11-
13]. The efficient and sustainable production methods can be
recognized to build the reliable agriculture strategies.
Machine learning is used for crop prediction in many
techniques. Intelligent techniques have been suggested by
various researchers and have given accurate results as well
[14]. These tests have been done on small number of crops.
Gonzalez-Sanchez et al.’s study [15] included data from a
large number of crops. Accuracy of different learning methods
was tested in this analysis and some techniques were proposed
to predict crop growth [16]. The Support vector technique is
more accurate and precise as compared to ARIMA. It is very
easy to implement as compared to statistical approaches
[17],[30].
In [18] the author shows that factors like temperature of
pacific and Indian ocean, monsoon rainfall and pressure of sea
level have direct impact on growth of agriculture products in
Pakistan. Moreover, the results prove the production rate and
amount of monsoon rainfall, all over Pakistan remain
constant, except few cases. Machine learning algorithm U
ChooBoost is used to accommodate idea of PA [19]. Lots of
digital information obtained from farm sensors needs to be
manipulated. Knowledge mining uses this supervised learning
algorithm (UChooBoost) [20]. UChoo classifier ensembled
over machine learning. PA votes are assigned with specific
weights and highest weighted votes are combined to check the
performance. Which boost up the performance. The extensive
experimentation is carried out over extended data expression.
Results of many assumptions can also be tested with its help
which will help in improvement of algorithm performance.
Information which is collected by farms sensors is used by
artificial neural networks for predication of production rate of
crops. Parameters like temperature, soil, rainfall, pressure and
humidity in this information. These parameters and their
effects on crop growth are discussed and the results evaluated
[21].
In [22] the authors develop a system to detect animal growth
in the livestock. It remains successful and very effective for
reducing cost, efficient source of energy, and robustness. The
Mobile Monitoring System based on RFID to hold the cattle
efficiently by extracting dynamic information about their
locality and behavior by wireless network. Their behavior is
analyzed. The used IoT sensors
with embedded system. The measure many parameters for
further analysis through this IOT-based system in a smart way.
In [23] cattle are monitored in the fields. The system is
developed over Zigbee and WSN. They effectively used WSN
for identifying the locality of the cattle.
A team of researchers [24] develop disease detection system
with “infrared thermography”. They efficaciously detect
disease over foot and mouth of livestock. They used infrared
thermography for screening the disease.
Scientist [25] recorded the heartbeat of the cattle. For the
purpose they used polar sport tester (PST) monitoring
technique for cattle. They monitor animal for their
temperature and heartbeat to identify the diseases.
In [26] the authors developed a real time system by using the
technology of smart mobiles and Bluetooth system with IOT.
They recorded the parameters like temperature and heartbeat
rate continuously.
The health monitor devices are not available in Pakistan
market veterinary doctors monitor it manually. [27],[29].
But sensor for soil temperature and humidity at different
levels are available which are accessible for implementation
and testing for our framework.
3. PROPOSED SYSTEM
In this manuscript we proposed novel framework for smart
framing with the federation of Machine learning prediction
system and real-time data collections by sensors, Raspberry
pi3 and IOT technology. We take agriculture practices as use
case. The data is collected in DHT11. Raspberry pi is stored
on cluster by employed the Dstream and txt splitting and
cleaning techniques. Once the data is stored, with federated
queries it can be accessed for planning and implementation of
the specific agriculture practice.
The temperature and humidity are main features sensor
DHT11 to ensure sustainability. The data being processed by
Raspberry pi will be updated continuously and passed to
streamed cluster. By data analytics the facts and figure can be
presented to the user. The recommendation is also made for a
specific condition.
Advantages of proposed system
• The temperature and humidity of a particular place to a
particular depth for the specific crop can be record
automatically with adjustment.
• The raspberry pi 3 requires very small processing space.
• It is very useful for predicting and forecasting the upcoming
climate risks
• Accessible from anywhere, efficient, user friendly.
3.1 Framework

Abdul Rehman et al., International Journal of Emerging Trends in Engineering Research, 8(9), September 2020, 6482 – 6487
6484
The monitoring node is raspberry pi3 model b monitor each
node and collect real-time data in the form streams which are
converted to Discretized Stream (DStream) which is
continuous form of Resilient Distributed Datasets (RDDs).
With spark core functionality and SparkSQL we become
enable to answer the federated query which enables the system
to make some recommendations.
For getting the continuous and sustainable real-time data the
sensor needed to be connected to the raspberry pi3 model b
with jumper wires as in figure no 3.
Figure 1: Framework of smart Farming
T1,T2,T3,T4,T5,T6,T7 are the series of times when the data
is broadcasted. And W1, W2, W3 are the dstream windows.
Which can be stored and can queried by SparKSQL. SparkML
is the module of the spark which is used to implement different
classifier of machine learning to extract the continuous and
discreate, diversified agri-data. The data is available in
heterogenous formats like text, web data, and CSV etc. the
required information is extracted and build rule of farming for
recommendations as shown in figure 1. When real information
is matched, and it is uneven it makes the recommendations.
3.2 Real-time DATA
The data is captured as mentioned in fig 1. For sensors in same
way it also can be collected from satellites and other machines
etc. In this paper, we present a simple framework for smart
farming to highlight the new research directions. The data
collect about the climate change and we made predictions
accordingly with machine learning what it will have effects
over plants.
3.3 Connectivity
As first step we can access raspberry pi interface on laptop
by following steps
• Wireless WIFI connection is published and shared
• Through SSH the GUI of raspberry pi is accessible by
USB /ethernet cable consoles.
• For identify the default IP address advanced IP scanner
can be used.
• For GUI Xming server and for ssh Xshell can be used to
view the SSH and GUI interface.
•
When it became accessible it shows the login, the default login
and password, which are mentioned on it. We can connect to
the terminal by giving the default password as shown in figure
2.
Figure 2: Password Screen

Abdul Rehman et al., International Journal of Emerging Trends in Engineering Research, 8(9), September 2020, 6482 – 6487
6485
Now the commands ssh with -X on XShell or Xmanger. We
can visualize the GUI interface Connections. The way of connection of Raspberry pi with DHT11 sensor using project
board and jumper wires are shown in figure 3.
Figure 3: Outer Interface Connection Setting
Let take an example of sugarcane which has different growth
stages such as
Germination of duration: 15 - 45 days,
Tillering of duration: 92 days,
vegetative of duration: 93 days
maturing normally duration: 60-75 days.
The rule for temperature is as followed for different stages.
Table 1: Temperature from IoT Devices
Stages
Optimal Temperature
Sowing
25-30 C
Germination
28 -30 C
Tillering
< 30 C
Vegetative
28-38C
Maturing
10C
For the rule generated from the data by machine learning,
when the value of temperature form real-time system become
worst, the system generate serious alert and also made
recommendation that how, the temperature can be lessened or
increased to get the optimal value. Is the irrigation required.
When the temperature is greater or less then the optimal value
but not to worry, the recommendation system sends warnings.
The data has been taken from the sensors using IoT setup and
shown in table 1 and table2. Here we show the sample
readings for temperature and humidity. We can see that
Temperature and humidity are changes day to day in
surrounding environment. From these three different samples
calculate minimum, maximum and average values of both
temperature and humidity. Here the DStream windows is the
day when the day reading has conflict with the rule for the
case sugarcane, the system generates alerts and made
recommendations with rules and suggestions
From the table 1 and table2, we take readings of temperature
and humidity for three samples.
Table 2: Humidity from IoT Device
.
4. CONCLUSION
In the manuscript we elaborate the novel framework of
recommendations system for smart and precision farming. We
try to ensure its sustainability be federating different
heterogeneous data from diversified resources. We conduct
analysis and made rules for making recommendations and
suggestions. We also predict the upcoming risks on the basis
of available data. The data includes the various parameters
including weather forecast, soil conditions, atmospheric
conditions, irrigation plan, cropping pattern information, pest
scouting and control policies by Government, fertilizers doses
and applications for various crops and crop yields etc. we
used the partial for this prototype in future we use the whole
data for big and complete solution of smart farming. The
implementations of real-time protype by using DHT11 sensor
show the new way of research to be get informed about the
field about the atmospheric and soil conditions of data on time.
We effectively used the big data technologies like Hadoop
ecosystem with spark to predict and suggest the plan for the
agricultural practices.
Date
Sample
1
Sample
2
Sample
3
Ma
x
Mi
n
GM
29.04.202
0
32
34
32
34
33
32.65
3
30.04.202
0
28
29
33
33
28
29.92
4
31.04.202
0
26
30
33
33
26
29.52
5

Abdul Rehman et al., International Journal of Emerging Trends in Engineering Research, 8(9), September 2020, 6482 – 6487
6486
REFERENCES
[1] C. Abrar and A. Nadia. Climate-Smart Agriculture in
Pakistan, CIAT / World Bank
https://cgspace.cgiar.org/handle/10568/83340, 2017
M. Kaur, G. Heena and H. Kundra. Data Mining in
Agriculture on Crop Price Prediction Techniques and
Applications, International Journal of Computer
Applications, Vol. 99, no.12, pp. 1-3, August 2014.
[2] S. Hussain, F. Kabeer, W. Ali and K. Jamshed. A New
Framework to Integrate Wireless Sensor Networks with
Cloud Computing, in Proc. of IEEE Aerospace Conference,
USA, 2013, pp. 2-9.
[4] F. X. Diebold. On the Origin(s) and Development of the
Term “Big Data, Penn Institute for Economic Research,
Pennsylvania, 2012, pp-12-37.
[5] D. Laney. 3-D Data Management: Controlling Data
Volume, Velocity and Variety, META Group Research Note,
Vol. 6, no. 70, pp.1, 2001.
[6] Apache Spark: https://spark.apache.org/
[7] V. C. Patil, K. A. A1-Gaadi, D. P. Biradar and M.
Rangaswamy. Internet Of Things (IoT) and Cloud
Computing for Agriculture: An Overview, in Proc. of agro-
informatics and precision agriculture (AIPA), India, 2012, pp.
292-296.
[8] E. Khandakar, A. Unayes and M. A. Gregory. Integrating
Wireless Sensor Networks with Cloud Computing, in Proc.
of 7th International Conference on Mobile Ad-hoc and Sensor
Networks (MSN), Beijing, 2011, pp.16-18.
[9] D. S. Sawaitul, K.P. Wagh and P. N. Chatur.
Classification and Prediction of Future Weather by using
Back Propagation Algorithm- An Approach, International
Journal of Emerging Technology and Advanced Engineering,
Vol. 2, no. 1, pp. 110-113, January 2012.
[10] I. Jagielska, C. Mattehews and T. Whitfort. An
investigation into the application of neural networks, fuzzy
logic, genetic algorithms, and rough sets to automated
knowledge acquisition for classification problems, Neuro
computing, Vol. 24, no. 1-3, pp.37-54, 2012.
[11] D. Ramesh and B. V. Vardhan. Analysis of Crop Yield
Prediction Using Data Mining Techniques, International
Journal of Research in Engineering and Technology, Vol. 4,
no. 1, pp. 47-473, January 2015.
[12] S. Veenadhari, B. Misra and C. D. Singh. Data Mining
Techniques for Predicting Crop Productivity – A Review
Article, International Journal of Computer Science and
Technology IJCST, Vol. 2, no. 1, pp. 90-100, March 2011.
[13] W. Duncan, K. Rabah. Environmental Conditions’ Big
Data Management and Cloud Computing Analytics for
Sustainable Agriculture, World Journal of Computer
Application and Technology, Vol.2, no.3, pp. 73-81, 2014.
[14] R. Piyare, S. Park, S. Y. Maeng, S. H. Chan, S. G. Choi,
H. S. Choi and S. R. Lee. Integrating Wireless Sensor
Network into Cloud Services for Real-time Data
Collection, in Proc. of International Conference on ICT
Convergence, Jeju, pp. 752-756, 2013.
[15] V.Rajesh, J. M. Gnanasekar, R. S. Ponmagal and P.
Anbalagan. Integration of Wireless Sensor Network with
Cloud, in Proc. of IEEE International Conference on Recent
Trends in Information, Telecommunication and Computing,
Kochi, pp. 321-323, 2010.
[16] R. Kumar, M. P. Singh, P. Kumar and J.P. Singh. Crop
Selection Method to Maximize Crop Yield Rate using
Machine Learning Technique. In Proc. of IEEE
International Conference on Smart Technologies and
Management for Computing, Communication, Controls,
Energy and Materials ICSTM, India, 2015, pp. 138-145.
[17] Z. Peng, Y. Zheng and S. Hamlin. A Novel Architecture
Based on Cloud Computing for Wireless Sensor Network,
in Proc. of 2nd International Conference on Computer
Science and Electronics Engineering ICCSEE, pp 472-475,
2013.
[18] R. Sudarsan, V. Rao, N. Satyanarayana and V. L.
Prasanna. Generic Integrated Secured WSN-Cloud
Computing for U-Life Care, International Journal of
Engineering Science and Advanced Technology, Vol. 2, no. 4,
pp 897-907, 2011.
[19] M. Firdhous, O. Ghazali and S. Hassan. Trust
Management in Cloud Computing: A Critical Review,
International Journal on Advances in ICT for Emerging
Regions, Vol. 4, no. 2, pp. 24-36, 2011.
[20] V. Rajesh, O. Pandithurai and S. Magesh. Wireless
sensor data on cloud, in Proc. of IEEE International
Conference on Communication, Control and Computing
Technologies, Ramanathapuram, 2010, pp. 476-481.
[21] Y. Rupika, J. Rathod and N. Vaishnavi. Big data meets
small sensors in precision agriculture, International
Journal of Computer Applications, Vol. 975, pp. 8887, 2015.
[22] A. Lounis, H. Abdelkrim, A. Bouabdallah and Y. Challal.
Secure and Scalable Cloud-based Architecture for e-
Health Wireless sensor networks, in Proc. of 21st
International Conference on Computer Communications and
Networks (ICCCN), Munich, 2012, pp 1-7.
[23] P. Langendoerfer, K. Piotrowski, M. Diaz and B. Rubio,
Distributed Shared Memory as an Approach for
Integrating WSNs and Cloud Computing, in Proc. of 5th
International Conference on New Technologies, Mobility and
Security, Istanbul, 2012, pp 1-6.
[24] C. Srimathi, P. Soo-Hyun, N. Rajesh. Proposed
framework for underwater sensor Cloud for
environmental Monitoring, in Proc. of 5th International
Conference on Ubiquitous and Future Networks, Da Nang,
2013, pp. 104-109.
[25] F. Tongrang, Z. Xuan and G. Feng. Cloud Storage
Solution for WSN Based on Internet Innovation Union,
International Journal of Database Theory and Application,
Vol. 22, pp. 164-169, 2013.
[26] S. Rajeswari, K. Suthendran and K, Rajakumar. A Smart
Agricultural Model by Integrating IoT, Mobile and Cloud-
based Big Data Analytics, International Journal of Pure and
Applied Mathematics, Vol. 118, no. 8, pp. 365-370, 2018.
[27] S. Rajeswari, K. Suthendran, K. Rajakumar and S.
Arumugam. An Overview of the MapReduce Model,
Lecture Notes in Computer Science, Vol. 10398, pp. 312-317,
2017.

Abdul Rehman et al., International Journal of Emerging Trends in Engineering Research, 8(9), September 2020, 6482 – 6487
6487
[28] M. Ongayev, Z. Sultanova, S. Denizbayev, G. Ozhanov
and S. Abisheva. Engineering and Process Infrastructure
of the Agro-Industrial Complex, International Journal of
Emerging Trends in Engineering Research, Vol. 7, no. 12, pp.
879-885, 2019.
[29] S. Aleksander, A. Konev, T. Kosachenko and D. Dudkin.
Threat Model for IoT Systems on the Example of
OpenUNB Protocol, International Journal of Emerging
Trends in Engineering Research, Vol. 7, no. 9, pp. 834-290,
2019.
[30] L. Boggula, B. Navyasri. Energy Efficient Routing
Mechanism for Harsh Environment in Wireless Sensor
Networks, International Journal of Emerging Trends in
Engineering Research, Vol. 7, no. 9, pp. 234-238, 2019.
[31] Apache kafka: https://kafka.apache.org/