Professional Documents
Culture Documents
net/publication/309614330
CITATIONS READS
16 1,256
2 authors, including:
Ravichandran Giritharan
Tamilnadu Police
8 PUBLICATIONS 29 CITATIONS
SEE PROFILE
Some of the authors of this publication are also working on these related projects:
All content following this page was uploaded by Ravichandran Giritharan on 15 February 2019.
I. INTRODUCTION
India is an Agricultural Nation. Almost majority of the
Indians traditional occupation is agriculture. Many of them
worship their land as the god. It is one of the prominent
factor that decides the economy of our country. Government
is also facilitating farmers with many features such as subsidy
in fertilizers, cancellation of their interest for loans etc. But
today the farmers are adversely affected by the crop loss.
Now a days there are variety of reason for the crop loss.
One of the most important factor is that there is no proper
literacy for farmers about the crops, fertilizers and productivity.
Also due these factors, the crop loss has increased and hence
the famers suicide has increased and the employment status in
Agriculture field has gradually decreased as shown Figure 2[1].
The drop in employment sector in the agricultural field is
shown in Figure 1[2].
Hence to aid them help in the technological point of view, a
prediction system is proposed that could intake the various
parameters that decide the productivity[3].
Figure 2. Report on Farmer’s suicide rates (Courtesy: The Hindustan Times)
The ANN has played a crucial role in developing various b. Hidden Layer
models. This article uses the most prominently used ANN, c. Output Layer
Feedforward Back propagation network.
The nodes are present in input layer as well as the output
The basic principle behind the operation of the proposed layer that corresponds to the variables at the input and output.
system is to take the inputs as various parameters that decide Here across the weighted connections the data traverses
the productivity, process them based on the algorithm provided between these nodes.
by ANN, and predict the suitable crop for the land. Also the
system suggests some fertilizers that could be used to improve In this particular model of propagation mechanism, each
the fertility. It also provides the Status of productivity of a node is excited by the previous node, and a weighted sum of all
particular crop in a particular soil. its inputs, t is calculated as follows:
B. Feedforward Backpropogation
One of the most commonly used algorithms or networks in
Artificial Neural Network is Feedforward Back Propagation
Network.
In this particular context, the above network is used
together to implement the Artificial Neural Network.
The neural process and recall patterns is done through the
Feedforward algorithm. The training to the proposed neural
network is given using the back propagation network. The
backpropogation is a supervised training methodology[11]- Figure 3. Layer connection of Fedforward back propogation network
[13].
III. DESIGN FLOW Step 5: Since the system is to be developed as the Android
The process of developing the proposed system involves application, the Matlab code of the system is extracted.
the following process: Step 6: The front panel of the Application is developed using
1. Data collection/ Preparation Eclipse IDE, and the Java program written at the backend.
2. Build the Prediction Model Step 7: Finally extract the application in the APK(Android
3. Classification Package Kit) format, so that it can be installed in a smart phone
4. Fertilizer suggestion for respective crop. and utilized.
All the details regarding the crops are collected from the TABLE I. DATABASE CONSIDERED FOR PREDICTION SYSTEM
Department of Agriculture and Science, Annamalai University
and Kishore Vigyan Agriculutral University which is shown in Rainf
Table I. pH N P K Depth T Crop
all
26- 750-
IV. BUILDING THE MODEL 7.2-8.4 110 45 50 20-30
32 1200
Cotton
The building of the Prediction and analyzing model using 25- 700- Sugar
6.3-7.6 175 100 100 40-60
Artificial Neural Network is divided into the following steps. 50 1200 Cane
24- 850-
Step 1, It is necessary to combine many different 5.8-8.5 85 35 45 5-20
29 100
Jowar
Artificial Neural Network (ANN) Prototypes, to find out the 29- 350-
optimal configuration. Hence certain algorithms were used for 7.2-8.6 50 25 20 15-25
32 750
Bajra
training this particular model such as Silva and Almeida’s 26- 750-
algorithm, Delta-bar-delta, Rprop, The Dynamic Adaption 6.5-7.8 25 70 20 15-20 Soybeans
34 1000
algorithm, Quickprop, QRprop[5],[7]-[9]. 14- 400-
7.1-8.4 90 25 10 20-50 Corn
28 600
Step 2, Trial and error method is used to decide the number
of hidden layers. Also certain value of training parameters are 15- 50-
6.2-8.6 100 50 50 15-25 Rice
24 200
obtained through the same method. A very careful observation
20- 800-
is needed to select the number of hidden layers. 5.4-8.5 110 50 50 5-20
25 1400
Wheat
TRANSFER FUNTION:
---------------------------------------------------------------------------
{tansig, tansig}
---------------------------------------------------------------------------
TRAINING ALGORITHM:
---------------------------------------------------------------------------
trainlm(Back Propagation Algorithm)
---------------------------------------------------------------------------
LEARNING:
---------------------------------------------------------------------------
Figure 4. Flow chart of prediction model
learngdm
---------------------------------------------------------------------------
Based on the above functions the ANN manipulates the input VII. RESULTS AND DISCUSSION
data and processes the information by linear network at the input
level. Then the information is propagated through the hidden
layers, where the sigmoidal function is applied to the data
processed and the desired result is produced.
Here following functions are used:
a) Sigmoidal function - Hidden and Output layers.
b) Linear function - Input layer.
V. SUGGESTION OF FERTILIZERS
The main objective of the proposed system is to suggest
crops to the farmer for his land. But there is an alternate case
in which a farmer want to use his own desired crop into his
field. In such controversial situation the proposed system has
an added advantage of suggesting the fertilizer for his land for
his desired crop. Also it says, whether the desired crop of the
farmer, can yield productivity or not.
Here Nitrogen, Phosphor and Potassium are the three
basic important minerals for a crop growth and hence the
fertilizer suggestion is based on these three values. If there
is the optimum availability of these basic nutrition in the soil, then
no fertilizers are required. When there occurs the deficiency
of nutrients then the fertilizers are suggested based on the
Table II.
Rain Crop
pH N P K Depth T
fall Predicted
80
this can further be improved by developing the predicting
70 system as a separate module using Raspberry Pi.
60 55
ACKNOWLEDGMENT
50 45
The authors would like to thank their family and
40 35 management of E.G.S. Pillay Engineering College for their
25 kind support.
30
20 REFERENCES
10 [1] Kunal Purohit. (Jul 21, 2015 ), " 1 in 3 farmer suicides in Vidarbha over
Rs.10,000debt:Study"[Online].Available:https://commons.wikimedia.org/wiki
0 /File:1951_to_2013_Trend_Chart_of_Sector_Share_of_Total_GDP_for_each
_year,_India.png.
10 20 30 40 50 60 70 80 90 100 [2] M Tracy Hunter, (18 July 2014), "File:1951 to 2013 Trend Chart of
Sector Share of Total GDP for each year, India.png"[Online],
Number of Hidden Layers Available:http://www.hindustantimes.com/mumbai/1-in-3-farmer-suicides-in-
vidarbha-over-rs-10-000-debt-study/story-6g5e4B92yLTrAPohoWlmNK.ht
[3] Snehal S. Dahikar, Sandeep V. Rode, Pramod Deshmukh, "An Artificial
Figure 9. System predicting the productivity status for the desired crop and Neural Network Approach for Agricultural Crop Yield Prediction Based on
suggests fertilizers to improve the productivity. Various Parameters", January 2015,vol 4, pp. 94-98.
[4] A,K Jain, J Mao and K,M Mohiuddin "Artificial Neural Networks: A
Here as discussed above the number of Hidden Layers are Tutorial", IEEE Computer, 1996, vol. 29, no. 3, pp.56 -63.
[5] A Tang, C., Quek, G. Ng, GA-TSKfnn; "Parameters tuning of fuzzy
selected by trial and error method. The Figure 9 shows the neural network using genetic algorithms", Expert Systems with Applications,
performance of the system for different number of Hidden 2005, pp. 769-781. "
Layers. From the above figure, we can observe the tradeoff [6] H. Ozturk, H. Ceylan, Q.E. Canyurt, A. Hepbasli, "Electricity estimation
between performance and complexity, with respect to the using genetic algorithm approach: a case study of Turkey", Energy, 2003,
Number of Hidden Layers. pp.1003-1012.
[7] F. Cavallaro, "Fuzzy TOPSIS approach for assessing thermal-energy
With increase in the Number of Hidden Layers, the storage in concentrated solar power (CSP) systems," Applied Energy. Oxford,
Performance increases along with the complexity. Hence to 2010, vol. 87, pp. 496-503.
[8] T. C. Chu and Y. C. Lin, "An interval arithmetic based fuzzy TOPSIS
resolve the tradeoff between these parameters, the optimum model," Expert Systems with Applications. Oxford, 2009, vol. 36, pp. 10870-
number of Hidden Layers is chosen by Trial and Error method, 10876.
and the obtained value is 50 at which the performance of the [9] Y. M. Wang and T. M. S. Elhag, "Fuzzy TOPSIS method based on alpha
system is 70%. level sets with an application to bridge risk assessment", Expert Systems with
Applications. Oxford, 2006, vol. 31, pp. 309-319.
VIII. CONCLUSION [10] Trigg HCE, "An Investigation of Methods to Enhance the Performance
of Artificial Neural Networks Used to estimate ICU Outcomes," MScE
From the above results it is clearly identified that the Thesis, U. of New Brunswick, January 1997.
[11] G.-B. Huang, "Learning capability and storage capacity of two-hidden-
proposed system functions properly on the input data, layer feedforward networks", IEEE Trans. Neural Networks, 2003 vol. 14,
manipulates it and provides the desirable output. The Matlab no.2, pp. 274-281, DOI: 10.1109/TNN.2003.809401
ANN Toolbox is the software domain in which the prediction [12] Sperduiti and A. Starita, "Speed up learning and network optimization
system is built and then imported to Android platform. with extended backpropagation", Neural Networks, 1993, vol. 6, pp. 365-383.
[13] G.-B. Huang and H. A. Babri, "Upper bounds on the number of hidden
Also the system is tested with some standard inputs and neurons in feedforward networks with arbitrary bounded nonlinear activation
verified for its results. The accuracy of the system is very high functions", IEEE Trans. Neural Networks, 1998, vol. 9, no. 1, pp. 224-229.
as discussed in the Results and Discussion session. One of the [14] Holger Schulze, Achim Spiller, "The Use of Artificial Neural Networks
as Information Systems in the Food Chain Management:An Empirical
most advantageous factor is that this particular system is there Comparison of Methods, Computers in Agriculture and Natural Resources",
is no cost for this particular system, and it can be installed in 4th World Congress Conference, Proceedings of the 24-26 July 2006, pp. 23-
any smart phones, which makes this system more comfortable 25, DOI: 10.13031/2013.21967
for usage. Also the farmers are accessible to these systems, [15] Yumei Shi, "Grey Model of Total Power of Agricultural Machinery of
and hence they would be properly educated regarding which Construction Corps in Xinjiang Autonomous Region", Agricultural
Machinaztion in Xinjiang, 2006, pp.31-32.
crop to grow, and which not to grow. [16] B.A. Smith et al, "Artificial Neural Networks for Automated Year round
The future work of this involves building this particular Temperature Prediction", Transaction on Computers and Electronics in
Agriculture, 2009, pp. 52–56.
application in the regional languages, so that it would be more [17] B. Ji et al, "Artificial neural networks for rice yield prediction in
comfortable for farmers as many of them are illiterates. mountainous regions", Journal of Agricultural Science, 2007, pp. 249–261.
Further it can be improved by adding certain add-ons such as