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

net/publication/322316222

Development and Analysis of Artificial Neural Network Models for Rainfall

Prediction by Using Time-Series Data

Article  in  International Journal of Intelligent Systems and Applications · January 2018

DOI: 10.5815/ijisa.2018.01.03

CITATIONS READS

27 1,244

4 authors, including:

Hemant Kumar Soni Sanjiv Sharma

AMITY UNIVERSITY MADHYA PRADESH Madhav Institute of Technology & Science Gwalior
15 PUBLICATIONS   61 CITATIONS    18 PUBLICATIONS   87 CITATIONS   

SEE PROFILE SEE PROFILE

Some of the authors of this publication are also working on these related projects:

International Journal View project

All content following this page was uploaded by Hemant Kumar Soni on 10 January 2018.

The user has requested enhancement of the downloaded file.

I.J. Intelligent Systems and Applications, 2018, 1, 16-23
Published Online January 2018 in MECS (http://www.mecs-press.org/)
DOI: 10.5815/ijisa.2018.01.03

Development and Analysis of Artificial Neural

Network Models for Rainfall Prediction by Using
Time-Series Data
Neelam Mishra
Department of Computer Science and Engineering, NRI College of Engineering and Management,
Gwalior, Madhya Pradesh, India

Hemant Kumar Soni

Department of Computer Science and Engineering, Amity School of Engineering and Technology,
Amity University, Gwalior, Madhya Pradesh, India
E-mail: [email protected]

Sanjiv Sharma
Department of Computer Science and Engineering, Madhav Institute of Technology and Science,
Gwalior, Madhya Pradesh, India

A K Upadhyay
Department of Computer Science and Engineering, Amity School of Engineering and Technology,
Amity University, Gwalior, Madhya Pradesh, India

Received: 19 May 2017; Accepted: 21 July 2017; Published: 08 January 2018

Abstract—Time Series data is large in volume, highly forecasting will help in evaluating drought and flooding
dimensional and continuous updating. Time series data situations in advance. Therefore, it is significant to have a
analysis for forecasting, is one of the most important perfect model for rainfall forecasting.
aspects of the practical usage. Accurate rainfall The country like India, where most of the farmers
forecasting with the help of time series data analysis will depend on monsoon for their crop, it is required to be
help in evaluating drought and flooding situations in having advance knowledge of actual rainfall. Many states
advance. In this paper, Artificial Neural Network (ANN) of India also suffered from the flood whereas some state
technique has been used to develop one-month and two- is facing the problem of drought. In any of these two
month ahead forecasting models for rainfall prediction situations, it must require an accurate and efficient
using monthly rainfall data of Northern India. In these rainfall prediction model. This kind of rainfall forecasting
model, Feed Forward Neural Network (FFNN) using model will help in better handling worst situations
Back Propagation Algorithm and Levenberg- Marquardt generated due to flood or draught. Advance prediction of
training function has been used. The performance of both rainfall by this model also gives sufficient time to makes
the models has been assessed based on Regression adequate arrangements for saving lives, transportation,
Analysis, Mean Square Error (MSE) and Magnitude of procurement and supply of food and medicines.
Relative Error (MRE). Proposed ANN model showed Data mining is a set of techniques used to extract
optimistic results for both the models for forecasting and unknown pieces of information from the large database
found one month ahead forecasting model perform better repository. There are various data mining techniques
than two months ahead forecasting model. This paper also available to extract valuable and useful information from
gives some future directions for rainfall prediction and spatial, temporal, sequencing and time series data.
time series data analysis research. Time Series data is a part of temporal data. Time series
data generated from scientific data, financial applications,
Index Terms—Data Mining, Time series data analysis, weather data, GPS, Sensor Networks etc. Large in
Rainfall forecasting, Artificial Neural Network, Feed volume, highly dimensional and continuous updating is
Forward Neural Network. the nature of time series data [1]. Use of time series data
[2] in prediction, pattern identification, anomaly detection,
motif discovery, clustering, classification, segmentation
I. INTRODUCTION fetches the attention of data mining researchers.
In time series data analysis the patterns which existed
Rainfall is recognizing as one of the main important in past will appear in future also. This time series data
element of the hydrological process. Accurate rainfall analysis allows a model for a complex system where the

Copyright © 2018 MECS I.J. Intelligent Systems and Applications, 2018, 1, 16-23
Development and Analysis of Artificial Neural Network Models for Rainfall 17
Prediction by Using Time-Series Data

goal is to predict the system’s behavior without knowing produced improved results with low computation cost for
the functionality of the system [3]. seasonal rainfall forecast. In continuation with previous
There are various statistical techniques, data mining models, The ANN and Wavelets Decomposition model
approaches, artificial neural network and soft computing [12] for rainfall predictions, CART and C4.5 [13] for
techniques including particle swarm optimization used by hourly rainfall predictions were also developed by
the researcher to solve time series data analysis problem researchers for forecasting. Amongst all improved Naive
[4, 5]. Bayes Classifier for time series data analysis had
Artificial Neural Networks (ANN) is an extensively accuracy rate 90% on rain /no-rain classification in the
established technique for modeling complex nonlinear comparison of genetic algorithm with average
and dynamic systems. ANN is helpful in developing an classification technique[14,15].
appropriate model when physical process relationship is Awan et al. [16] proposed Back Propagation (BP) and
not clear or where the nature of the event has chaotic Learning Vector Quantization (LVQ) technique for
properties. Although neural network required prior rainfall prediction. Author claim their techniques perform
information about the system but it reduces the model's better in terms of accuracy, lead time and required fewer
reliance on this prior information. This removes the need resources. The also assert that LVQ takes less training
for an exact specification of the precise functional form time than Back Propagation. Du et al. [17] suggested
of the relationship that the model seeks to represent. In Immune Evolutionary Algorithm based on Back
this paper, an attempt has been made to develop a 1- propagation network. Authors said their model have
month and 2-month rainfall forecasting model for rainfall higher accuracy and better stability. The author also
prediction in North India by using ANN. claims their model is suitable for solving complicated
This paper is divided into 7 sections. First Section problems of optimization. Jin et al. [18] proposed Particle
cover introduction, section two discuss the related work Swarm Optimization (PSO) - Neural Network ensemble
in this domain. The third section describes objective and prediction model for monthly mean rainfall. Authors
significance of the study. Section forth explain the data compared their result with traditional linear statistical
set used. The fifth section describes ANN model forecast method and claim superior predicting capability
development and section 6 cover result and discussion. and enhanced generalization capacity. Suhartono et al.
Paper end with a conclusion, limitations and future scope [19] implemented ensemble method based on ANFIS and
by section 7. ARIMA for monthly rainfall prediction. Decision Tree
Method using SLIQ [20] generated classification rule for
rainfall prediction with 72.3% accuracy. Feed-forward
II. RELATED WORK NN based NAR model [21] for forecasting time series
was also an important outcome for forecasting.
There are various methodologies used for time series
Genetic Algorith [22] and Artificial Neural Network
data analysis in which implementation of Neural Network [23, 24,25,26], Multi-Layer Perceptron (MLP),
(NN) is a popular choice for the researchers. Functional Link Artificial Neural Network (FLANN) and
Various methodologies have been used for time series
Legendre Polynomial Equation (LPE) [27], Multiple
data analysis by researchers in which performance of
Linear Regression (MLR) techniques [28] were
Neural Network (NN) is accepted and best option for the introduced for Rainfall prediction.
researchers amongst the different techniques like
Generalized Regression, Pearson Coefficient Technique,
Fuzzy Inference System, Focused Time Daley Neural
III. OBJECTIVE AND SIGNIFICANCE OF THE STUDY
Network etc. Various work has been carried out by
researchers on time series data analysis. In this regard, the The prime objective of this study is to analyze rainfall
model was developed by using Generalized Regression pattern in North India using appropriate time series
Neural Network [6] and compared with the model based methods for the period 1871 to 2012 ( 141 years) on the
on the Back propagation Neural Network. The basis of data recorded at Indian Meteorological
Generalized Regression Neural Network yielded simple Department, Pune. The specific objectives are:
and stable structure, with an accurate prediction of annual
rainfall. Pearson Coefficient Technique [7] was compared 1. To develop a time series model for one-month and
with Regression Approach and found that the predicted two-month ahead rainfall forecasting
values are lie below computed value for time series data 2. To identify the best network structure for both the
analysis. Apart from above techniques, Fuzzy Inference models.
System [8] was observed as a good alternative method for 3. Compare the performance of one-month and two-
accurate prediction. Accurate results can also be obtained month ahead forecasting models.
by using Focused Time Daley Neural Network model [9].
The model’s predictions were on monthly, quarterly and Weather and climate over the earth are not constant
annually basis. NNARX and ANFIS [10] models were with time, they change on different time series ranging
also found accurate for rainfall-runoff prediction and from the geological to the diurnal through annual,
reliable forecasting. Model using Neuro-Fuzzy Neuron seasonal and intra-seasonal timescales. Rainfall varies
Technique [11] was designed and compared with with latitude, elevation, topography, seasons, distance
Dynamic Downscaling Model. It is found that model from the sea, and Coastal sea-surface temperature. Trends

Copyright © 2018 MECS I.J. Intelligent Systems and Applications, 2018, 1, 16-23
18 Development and Analysis of Artificial Neural Network Models for Rainfall
Prediction by Using Time-Series Data

in rainfall extremes have enormous implications. Flood or IV. ANN MODEL DEVELOPMENT
draught events cause significant damage to agriculture,
Artificial Neural Network (ANN) Model is developed
environment, infrastructure, social life and on GDP.
using MATLAB. In this model Feed Forward Neural
Therefore, a reliable rainfall forecasting and assessing
Network (FFNN) using Back Propagation algorithm and
behavior at the station, regional and national levels is
Levenberg- Marquardt training function has been used.
very important. The results of this research paper will
Trials are first conducted by randomly selecting the
hopefully be used:
number of processing elements at the end of the training,
on the basis of a mean square error, accuracy will be
1. For forecasting the pattern of rainfall in the study
predicted. As per the performance of model and resultant
area.
predictions, the selection of the processing elements
2. To provide information that would be helpful for
carefully narrowed down. This ANN model generates a
decision makers in formulating policies to mitigate
very accurate prediction for the range of 5 to 40 neurons.
the problems of rainwater resources management,
This ANN model consists of one hidden layer and a
soil erosion, flooding, and drought.
varying number of processing elements or neurons. The
3. To provide information for the early warning
most optimal model obtained by using trial and error
system in the study area.
approach. The training parameters applicable for model
4. As a basis for further study in Northern India.
training are given in the Table-1.

Table 1. Model parameter values for Back Propagation Algorithm for all the models

Parameters used for Network Training

Network Type Feed Forward Neural Network with Back Propagation
Training Functions Used Levenberg-Marquardt (LM)
Adoption Learning Function learnGDM and learnGD
Performance Function Mean Square Error (MSE); Regression (R)

For Hidden Layer – Tansigmoid

Transfer Function
For Output Layer – Linear

No. of neurons used for hidden layer 5 to 40

This study focuses on the North India only, so only

V. DATA SET USED stations located in this area were selected, while the other
stations which are located outside were discarded. The
For this one-month and two-month ahead rainfall
data used as input and output variables for optimum
forecasting model development, a monthly time series model development are given in Table 2 below. Here
rainfall data of North India for the period 1871 to 2012 two models shall be developed, model M1 is for one-
( 141 years) were used. This time series data collected by
month ahead prediction and the other model M2, is for
Indian Meteorological Department, Pune. The data were
two-month ahead prediction. In both M1 and M2 models,
collected by various stations of different states of three input variables have been used which include
Northern India. The site was chosen due to the successive months of rainfall data. The output in both the
availability of relatively long series of meteorological
models is the predicted rainfall, which is one month and
data.
two month ahead values.
After analyzed data, some data was selected to train
ANN models, and the remaining was used as a testing set.

Table 2. Model Input / Output Variables

Model Input Variables Rainfall (mm) Output Variables Rainfall (mm)

R(t-2)
M1 R(t-1) R(t+1)
R(t)
R(t-2)
M2 R(t-1) R(t+2)
R(t)

number of neurons in the hidden layer is arranged in

VI. RESULTS AND DISCUSSIONS ascending order and their validation performance is
measured accordingly.
Network performance of both the models M1 and M2
are given in tables below. In Table 3 and Table 4, the

Copyright © 2018 MECS I.J. Intelligent Systems and Applications, 2018, 1, 16-23
Development and Analysis of Artificial Neural Network Models for Rainfall 19
Prediction by Using Time-Series Data

Table 3. Network Performance for Model M1 for one Month Ahead Forecasting

No. of Iteration
Neurons Used Model Performance Stopping Criteria Network Configuration
(Epochs)
5 120914.77 12 Val. Stopped/Max. iterations 3-5-1
10 120332.90 14 Val. Stopped/Max. iterations 3-10-1
15 129833.71 22 Val. Stopped/Max. iterations 3-15-1
20 136442.98 8 Val. Stopped/Max. iterations 3-20-1
25 (best) 112461.95 4 Val. Stopped/Max. iterations 3-25-1
30 152499.50 23 Val. Stopped/Max. iterations 3-30-1
35 103683.66 24 Val. Stopped/Max. iterations 3-35-1
40 122817.65 13 Val. Stopped/Max. iterations 3-40-1
45 (best) 132373.42 8 Val. Stopped/Max. iterations 3-45-1
50 154452.31 6 Val. Stopped/Max. iterations 3-50-1

Table 4. Network Performance for Model M2 for two Month Ahead Forecasting

No. of Iteration
Neurons Used Model Performance Stopping Criteria Network Configuration
(Epochs)
5 196653.42 34 Val. Stopped/Max. iterations 3-5-1
10 169111.91 44 Val. Stopped/Max. iterations 3-10.1
15 142875.66 29 Val. Stopped/Max. iterations 3-15-1
20 146643.47 12 Val. Stopped/Max. iterations 3-20-1
25 198612.44 11 Val. Stopped/Max. iterations 3-25-1
30 220510.21 8 Val. Stopped/Max. iterations 3-30-1
35 208755.18 5 Val. Stopped/Max. iterations 3-35-1
40 233804.28 10 Val. Stopped/Max. iterations 3-40-1
45 159821.84 13 Val. Stopped/Max. iterations 3-45-1
50 (best) 138643.47 12 Val. Stopped/Max. iterations 3-50-1

It is observed that for Model M1 and M2, the best after 12 and 9 epochs, the performance of training, testing
network structure is 3-25-1 and 3-50-1 respectively. It is and validation errors were stagnant. It means after epochs
also found out that the performance is not necessarily 12 and 9 there is no further improvement in the
improved even when the network error is low. performance of the network and the network seems to
It is very much clear from figure 1(a) and 1(b) that have saturated.

(a) (b)
Fig.1. Training of NN Model gauged by MS for M1 and M2 Models respectively.

(a) for N=45 (b) for N=50

Fig.2. Scatter Plot of R-Values for Training, Testing and Validating datasets for best-developed network for M1 and M2 Models Respectively

Copyright © 2018 MECS I.J. Intelligent Systems and Applications, 2018, 1, 16-23
20 Development and Analysis of Artificial Neural Network Models for Rainfall
Prediction by Using Time-Series Data

Linear regression analysis between the network parameters are in acceptable proximity with actual values.
response and the network output shown in figure 2. It can This representation, therefore, agrees with the conclusion
be inferred that NN model does the good mapping. In this that, high accuracy of prediction is attained by Neural
study, 15% data used for validation purpose, which is other Network Model after successful completion of training
than training data for the model. Thus, the performance of criteria i.e. with the value of MSE being within an
these machining conditions never experienced before by acceptable range as well as an agreeable performance
the neural network model. Hence, this mapping considers measure. Hence, from the results, it is inferred that the
to be true and representing the functional relationship. performance of the NN model is acceptable. Thus, can
Following figures 3 given below depicts the further be confirmed from the scatter plots shown in 4.
comparison between actual and simulated data for surface Further analysis of the observed and predicted values for
roughness. It is noticed that except for rare occasions, both M1 and M2 models on the basis of MRE values are
simulated surface roughness values for the designated shown in figures 5.

(a) (b)
Fig.3. Comparison of observed and predicted values by best M1 & M2 models

(a) (b)
Fig.4. Scatter Plot of Observed vs. Predicted Rainfall values for M1 & M2 Models

(a) (b)
Fig.5. Deviation of Predicted value from Observed value for M1& M2 Models

Computed Regression and MSE values for Model M1

A. The performance of M1 Model:
and Model M2 for different network structures are given
in Table 3 and Table 4 respectively. For the network As shown in Table 3, M1 Model gives the best result
identification purpose, as given in the last column, the in 3-25-1structure, with the least MSE value of
first number indicates the number of neurons in the input 112461.95 and the best regression values of 0.946 and
layer, middle represents neurons in the hidden layer and 0.948 for the validating and testing data sets respectively.
the last number indicates the number of neurons in the Although the regression values are slightly inferior for
output layer. training data in the comparison of validation data
regression value and testing data regression value. The
same has been depicted graphically in figure 6 given
below.

Copyright © 2018 MECS I.J. Intelligent Systems and Applications, 2018, 1, 16-23
Development and Analysis of Artificial Neural Network Models for Rainfall 21
Prediction by Using Time-Series Data

Table 5. Statistical Parameters for different network structures for M1 Model

Structure No. No. of Neurons Regression Values MSE

Training Val. Validating Val. Testing Val. All
1 3-5-1 0.9211 0.927 0.93 0.923
2 3-10-1 0.936 0.937 0.941 0.937
3 3-15-1 0.943 0.931 0.92 0.937
4 3-20-1 0.946 0.921 0.916 0.937
5 3-25-1 0.935 0.946 0.948 0.939
6 3-30-1 0.948 0.912 0.929 0.939
7 3-35-1 0.95 0.941 0.922 0.945
8 3-40-1 0.951 0.931 0.93 0.945
9 3-45-1 0.942 0.936 0.943 0.942
10 3-50-1 0.949 0.916 0.937 0.941

Fig.6. Graphical representation of Regression values for M1 Model

Fig.7. Graphical representation of Regression values for M2 Model

Further from table 3 and figure 3(b), it is observed that, but there is a slight fall in the values of R for testing and
in general, the R values for training are in general better validating datasets.
than testing and validating datasets, which shows that Further, a comparative analysis of M1 and M2 model
the network seems to have performed well for training based on tables 5 and 6 shows that M1 model for one-
datasets. Further, from the detailed analysis of figure 3, it month ahead rainfall, values have better R for training,
is observed that as the size of the network structure testing and validating datasets than an M2 model for
increases, the R-value for training dataset becomes better two- months ahead rainfall prediction.

Table 6. Statistical Parameters for different network structures for M2 Model

Structure No. of
Regression Values MSE
No. Neurons
Training Validating Testing
All
Value Value Value
1 3-5-1 0.897 0.886 0.903 0.896
2 3-10-1 0.901 0.904 0.884 0.900
3 3-15-1 0.901 0.904 0.905 0.902
4 3-20-1 0.907 0.924 0.849 0.901
5 3-25-1 0.912 0.879 0.892 0.904
6 3-30-1 0.908 0.876 0.875 0.899
7 3-35-1 0.905 0.889 0.88 0.899
8 3-40-1 0.905 0.856 0.909 0.899
9 3-45-1 0.909 0.894 0.901 0.905
10 3-50-1 0.913 0.869 0.910 0.906

Copyright © 2018 MECS I.J. Intelligent Systems and Applications, 2018, 1, 16-23
22 Development and Analysis of Artificial Neural Network Models for Rainfall
Prediction by Using Time-Series Data

parameters for better predictions.

B. The performance of M2 Model:
3. In a further study, the models may be developed
As shown in Table 4, M2 Model gives the best result in for the entire nation by using monthly, monsoon
3-50-1structure, with the least MSE value of 138643.47 and annual data of entire country.
and the best regression values of 0.913 and 0.910 for the 4. The effect, impact and relation to Global Warming
training and testing datasets respectively, with overall R- and climate change on rainfall may be a hot issue
value of 0.906 . Although the regression values are for future research.
slightly inferior for validation data in the comparison of 5. To improve the efficiency of proposed system, an
training data regression value and testing data regression online learning method for a neural network may
value. The same has been depicted graphically in figure 7. be used.

REFERENCES
VII. CONCLUSION, LIMITATIONS AND FUTURE WORK [1] Fu Tak-chung. ―A review on time series data mining‖.
Engineering Applications of Artificial Intelligence. vol. 24,
A. Conclusion pp 164–181, 2011.
[2] P. Esling, C. Agon. ―Time-Series data mining‖. ACM
In this study, 141 years rainfall records were analyzed,
Comput. Surv. vol. 45, 1, (12) ,pp. 1-34, 2012.
using data from Indian Meteorological Department, Pune [3] ―Time series analysis‖, Ramasubramanian V, IASRI
from various weather stations in the North India. Report. IASRI, Library Avenue, New Delhi, 2007
Artificial Neural Network (ANN) technique has been [4] Neelam Mishra, Hemant Kumar Soni, Sanjiv Sharma. ―A
used to develop one-month and two- month ahead Comprehensive Survey of Data Mining Techniques on
forecasting models for rainfall prediction using monthly Time Series Data for Rainfall Prediction‖. Journal of ICT
rainfall data. In these models, Feed Forward Neural Research and Application, in press.2017
Network (FFNN) using Back Propagation algorithm and [5] N. Filzah, M. Radzuan, Z. Othman, A. Abu Bakar.
Levenberg- Marquardt training function has been used. ―Uncertain Time Series in Weather Prediction‖. Procedia
Technology, vol 11, pp. 557 – 564, 2012.
The performance of both the models have been assessed [6] W. Zhi-liang, S. Hui-hua. ―Rainfall Prediction Using
based on Regression Analysis, Mean Square Error (MSE) Generalized Regression Neural Network: Case study
and Magnitude of Relative Error (MRE).Based on the Zhengzhou‖. International Conference on computational
overall results of the research, the following conclusions and Information Sciences. pp.1265-1268, 2010.
could be drawn: [7] M Kannan, S Prabhakaran, P. Ramachandran. ―Rainfall
Forecasting Using Data Mining Technique‖. International
1. Two models M1 and M2, for one month and two Journal of Engineering and Technology. Vol. 2(6), pp.
months ahead prediction of rainfall, having 397-401, 2010.
different input variables were trained and tested [8] K. Jesada, W. K. Wai, F. C. Che. ―Rainfall Prediction in
the Northeast Region of Thailand Using Modular Fuzzy
using ANN technique. Inference System". World Congress on Computational
2. In both the cases for M1 and M2 models, 3-25-1 Intelligence.,vol 10(15), pp. 136-141, 2012.
and 3-50-1 network structure model was found to [9] R Shamin M A, D Han, J Mathew. ―ANFIS and NNARX
be the best forecasting model. based Rainfall-Runoff Modeling‖. IEEE International
3. ANN has performed better for M1 model than M2 Conference on Systems Man and Cybernetics, pp. 1454-
model, showing that ANN technique has been able 1459, 2008.
to perform better for one month than two months [10] K. K. Htike, O. O. Khalifa. ―Rainfall Forecasting Models
ahead. Using Focused Time-delay Neural Networks‖.
4. The regression values for the M1 model have International Conference on Computer and
Communication Engineering. vol. 11(13), pp. 1-6, 2010.
shown better results than the M2 model.
[11] T. N. Castro, S. Francisco, JMB Alves, RST Pontes,
B. Limitations of the Study MBM Firmino, TM Pereira. ―Seasonal Rainfall Forest
using a Neo-Fuzzy Neuron Model. IEEE International
1. This study is based on the data of Northern part of Conference on Industrial Informatics (INDIN), pp. 694-
India, which does not cover the entire nation. 698, 2011.
2. Rainfall varies with latitude, elevation, topography, [12] W Phusakulkajorn, C Lursinsap, J Asavanant. ―Wavelet-
seasons, distance from the sea, and coastal Sea- Transform Based Artificial Neural Network for Daily
surface temperature. In these two models, these Rainfall Prediction in Southern Thailand‖. 9th
International Symposium on Communications and
constraints are not considered.
Information Technology, Icheon, pp. 432-437, 2009.
3. These models are for only one-month and two- [13] J. Soo-Yeon, S. Sharma, Y. Byunggu, J. D. Hyun.
month ahead prediction. It does not cover the ―Designing a Rule-Based Hourly Rainfall Prediction
entire monsoon period or for a complete year. Model‖. IEEE IRI, pp. 303-308, 2012.
[14] N.A. Charaniya, S.V. Dudul. ―Committee of Artificial
C. Future Work
Neural Networks for Monthly Rainfall Prediction using
1. An Alert and Decision Support System can be Wavelet transform‖. International Conference on
developed which can be helpful in flood or Business, Engineering and Industrial Applications, pp.
draught situation. 125-129, 2012.
[15] JNK Liu, BNL Li, TS Dillon. ―An Improved Naïve
2. A future study can be incorporated multiple

Copyright © 2018 MECS I.J. Intelligent Systems and Applications, 2018, 1, 16-23
Development and Analysis of Artificial Neural Network Models for Rainfall 23
Prediction by Using Time-Series Data

Bayesian Classifier Technique Coupled with a Novel Authors’ Profiles

Input Solution Method‖, IEEE Transactions on systems,
man, and Cybernetics – Part C: Applications and Reviews. Hemant Kumar Soni received M.Sc. in
Vol 31(2), pp.249-256, 2001. Computer Science from Jiwaji University,
[16] JA Awan, O. Maqbool. ―Application of Artificial Neural Gwalior, Madhya Pradesh, India in the year
Networks for Monsoon Rainfall Prediction‖. Sixth 1996 and M. Tech (IT) from Bundelkhand
International Conference on Emerging Technologies, pp. University, Jhansi, Uttar Pradesh, India in
27-32, 2010. the year 2006. He is pursuing Doctoral
[17] D. Jiaxing , Z. Bin , M. Shaohui. ―An application on the degree in Computer Science and
Immune Evolutionary Algorithm based on Back Engineering from Amity University Madhya
Propagation in the Rainfall Prediction‖. International Pradesh, Gwalior, India. He has 21 years of teaching experience
Conference on Computer Science and Electronics for UG and PG courses in Computer Science and presently
Engineering, pp. 313-317, 2012. working as Head of the Department of Computer Science and
[18] J. Long, H. Ying, Zhao , Hua-sheng. ―Ensemble Engineering at Amity University, Gwalior, Madhya Pradesh,
Prediction of Monthly Mean Rainfall with a Particle India. His research interest in Data Mining and Soft Computing.
Swarm Optimization – Neural Network Model‖. IEEE IRI, He published many research papers in National, International
pp. 287-294, 2012. Conferences and Scopus Indexed Journals. He is Reviewer of
[19] S. R. Faulina, D. A. Lusia, B. W.Otok, Sutikno and Heri many referred journals. He received a Best Paper Award in an
Kuswanto. ―Ensemble Method based on ANFIS-ARIMA International Conference and organized number of National
for Rainfall Prediction‖. International Conference on level events and conferences. He is a member of International
Statistics in Science, Business, and Engineering (ICSSBE), Association of Engineers, Hongkong, Universal Association of
pp. 1-4, 2012. Computer and Electronics Engineers (UACEE), The Institute of
[20] N.Prasad, P. Kumar , MM Naidu. ―An Approach to Research Engineers and Doctors, USA, Life Member of ISTE
Prediction of Precipitation Using Gini Index in SLIQ (Indian Society for Technical Education), India and Member of
Decision Tree‖. 4th International Conference on IAENG Society of Computer Science and Data Mining, Hong
Intelligent Systems, Modeling and Simulation, pp. 56-60, King.
2013.
[21] R. Adhikari, RK Agrawal. ―Forecasting strong seasonal
time series with Artificial Neural Network‖. Journal of
Scientific and Industrial Research, vol. 71, pp. 657-666, How to cite this paper: Neelam Mishra, Hemant Kumar Soni,
2012. Sanjiv Sharma, A K Upadhyay, "Development and Analysis of
[22] S. Singh,J. Gill, ―Temporal Weather Prediction using Artificial Neural Network Models for Rainfall Prediction by
Back Propagation based Genetic Algorithm Technique‖, Using Time-Series Data", International Journal of Intelligent
I.J. Intelligent Systems and Applications,vol.6 (12), pp. Systems and Applications(IJISA), Vol.10, No.1, pp.16-23, 2018.
55-61,2014. DOI: 10.5815/ijisa.2018.01.03
[23] B.M. Al-Maqaleh, A.A. Al-Mansoub and F.N. Al-Badani,
―Forecasting using Artificial Neural Network and
Statistics Models‖, I.J. Education and Management
Engineering, vol.3, pp. 20-32, 2016.
[24] V. Nourani, T.R. Khanghah, and A.H. Baghanam,
―Application of Entropy Concept for Input Selection of
Wavelet-ANN Based Rainfall-Runoff Modeling‖,
Journal of Environmental Informatics, vol. 26 (1),
pp.52-70, Sep. 2015.
[25] R.C. Deo, M. Sahin, ―Application of the Artificial
Neural Network model for prediction of monthly
Standardized Precipitation and Evapotranspiration Index
using hydrometeorological parameters and climate
indices in eastern Australia‖, Atmospheric Research, vol.
161, pp. 65-81, August 2015.
[26] M. Valipour, ―Optimization of neural networks for
precipitation analysis in a humid region to detect
drought and wet year alarms‖, Meteorological
Applications, vol. 23 (1), pp. 91-100, January 2016.
[27] SK Nanda, DP Tripathy, SK Nayak, S Mohapatra.
―Prediction of Rainfall in India Using Artificial
Intelligent‖, Systems and Applications.; vol.12, pp.1-22,
2013.
[28] N. Sethi, K. Garg. ―Exploiting Data Mining Technique for
Rainfall Prediction‖. International Journal of Computer
Science and Information Technologies. Vol. 5 (3),
pp.3982-3984, 2014.

Copyright © 2018 MECS I.J. Intelligent Systems and Applications, 2018, 1, 16-23

View publication stats