How to Decide the Best Fuzzy Model in ANFIS

w w w . i j m r e t . o r g I S S N : 2 4 5 6 - 5 6 2 8 Page 1
International Journal of Modern Research in Engineering and Technology (IJMRET)
www.ijmret.org Volume 3 Issue 2 ǁ February 2018.
How to Decide the Best Fuzzy Model in ANFIS
Mehmet YILMAZ1
1
(Harran University, Engineering Faculty, Geodesy and Photogrammetry Engineering Department, Geodesy
Division, 63100, Osmanbey Campus, Sanliurfa Turkey)
ABSTRACT : Geoid height determination is one of the major problems of geodesy because usage of satellite
techniques in geodesy isgetting increasing. Geoid heights can be determined using different methods according
to the available data. Soft computing methods such as Fuzzy logic and neural networks became so popular that
they are used to solve many engineering problems. Fuzzy logic theory and later developments in uncertainty
assessment have enabled us to develop more precise models for our requirements. In this study, How to
construct the best fuzzy model is examined. For this purpose, three different data sets were taken and two
different kinds (two inpust one output and three inputs one output) fuzzy model were formed for the calculation
of geoid heights in Istanbul (Turkey). The Fuzzy models results of these were compared with geoid heights
obtained by GPS/levelling methods. The fuzzy approximation models were tested on the test points.
KEYWORDS -Adaptive Network Based Fuzzy Inference Systems (ANFIS), Most suitable number of subsets,
fuzzy logic, input-output systems,geoid height
I. INTRODUCTION
The geoid is an equipotential (level) surface of the
earth's gravity field which coincides with mean sea
level (MSL) in the open oceans. Geoid is very
important for defining height because it provides a
meaningful reference frame for it. The importance
of accurately modelling the geoid has increased in
recent years with the usage of satellite positioning
systems such as the Global Positioning System
(GPS), GALILEO and GLONASS. GPS provides
height information relative to a best-fitting earth
ellipsoid rather than the geoid [1]. The relationship
between the geoid and the ellipsoid must be known
to convert ellipsoidal heights derived from GPS to
conventional (and more meaningful)
orthometricheights[2]. The situation is illustrated in
Fig. 1.
The relation between ellipsoid, geoid and geoid
height can be written following simple equation
H = h – N (1)
where h are ellipsoidal heights obtained from GPS
observations, H are orthometric heights derived for
example from spirit levelling and gravimetry, N are
geoid heights [3],[4],[5].
Geoid determination methods can be classified
according to the available data and the methods
used. These methods are: Astro-geodetic,
gravimetric, Global geopotential, remove restore
[6] and GPS/levelling. The GPS/levelling method
is now preferred because of the increase of GPS
measurements and its simple equations. The
GPS/levelling geoid height determination method
can be used for regional or local geoid
determinations. The geoid heights determined by
the GPS/levelling method can be also used as data
in geoid height determinations by polynomial
coefficients, fuzzy logic, artificial neural network
and finite element method.
Fig. 1: The geometrical relationship
between orthometric and ellipsoidal height and
geoid height [7].
II. ADAPTIVE NETWORK BASED
FUZZY INFERENCE SYSTEMS
Adaptive Network based Fuzzy Inference
Systems (ANFIS) are feed-forward adaptive
networks which are functionally equivalent to
fuzzy inference systems. The basic idea of ANFIS

can be described as follows: A fuzzy inference
system is typically designed by defining linguistic
input and output variables as well as an inference
rule base. However, the resulting system is just an
initial guess for an adequate model. Hence, its
premise and consequent parameters have to be
tuned based on the given data in order to optimise
the system performance. In ANFIS this step is
based on a supervised learning algorithm [8].For
simplicity, assume that the fuzzy inference system
under consideration has two inputs x and y and one
output f. Additionally, suppose that the rule base
contains two fuzzy if-then rules of the Takagi and
Sugeno’s type [9], [10], [11], [12], [13] as
Rule 1: If X is A1 and Y is B1; then f1= p1x + q1y
+ r1.
Rule 2: If X is A2 and Y is B2; then f2 = p2x + q2y
+ r2.
( In the Same way, fuzzy inference system under
consideration has three inputs x, y and z and one
output f. two fuzzy if-then rules of the Takagi and
Sugeno’s type as
Rule 1: If X is A1 and Y is B1 and Z is C1; then
f1= p1x + q1y +s1z + r1.
Rule 2: If X is A2 and Y is B2 and Z is C2; then f2
= p2x + q2y +s2z + r2)
The associated ANFIS architecture is illustrated in
Fig. 2 [14]. The functions of each layer can be
summarised such that In layer 1, inputs were
divided subspaces using selected membership
function. The membership can be chosen to be a
Gaussian shaped with the maximum value equal to
1 and minimum value equal to 0 such as, e.g., the
Gaussian function (shown in Fig. 3)[15].
Where, {σ,ci} is the premise parameter set. In layer
2, firing strength of a rule is calculated by
multiplying incoming signals. In layer 3, firing
strengths are normalised and in layer 4, consequent
parameters {pi, qi, ri} are determined and in layer 5,
final output is obtained by summing of all
incoming signals. The training algorithm, namely
ANFIS, was developed by Jang [1993]. Basically,
ANFIS takes the initial fuzzy model and tunes it by
means of a hybrid technique combining gradient
descent back-propagation and mean least-squares
optimization algorithms (see Fig. 4). At each
epoch, an error measure, usually defined as the sum
of the squared difference between actual and
desired output, is reduced. Training stops when
either the predefined epoch number or error rate is
obtained. The gradient descent algorithm is mainly
implemented to tune the non-linear premise
parameters while the basic function of the mean
least-squares is to optimize or adjust the linear
consequent parameters. [7], [8], [16], [17], [18],
[19], [20], [21], [22], [23], [24].
Fig. 2. A simple two inputs, two rules and a single
output of ANFIS structure [14]
Fig. 3: Physical meanings of the parameters in the
Gaussian membership function [15].
Fig. 5 shows an example of fuzzy partitioning of
the input space in case of two inputs. Each input is
divided by three subsets. So, the input space is
partitioned into nine fuzzy subspaces, leading to
nine fuzzy if-then rules in the ANFIS.
Fig. 4: ANFIS learning using hybrid technique
(LMS :Least mean square, RMSE :root mean
square error) [25]
[] RMSE
Gradiaent
descent
Premise
Parameters
(a, b, c)
Consequent
Parameters
(p,q, r)
LMS
Forward Pass
Backward Pass
Update
X
Y
A1
Π
Π
N
N
Z
1
w
2
w
1
w
2
w
11
fw
22
fw
A2
B1
B2
if X is A1 and Y is B1, then f1 = p1x + q1 + r1
if X is A2 and Y is B2, then f2 = p2x + q2 + r2
Layer(1) Layer(2) Layer(3) Layer(4) Layer(5)
Σ

Fig. 5: Corresponding fuzzy subspaces for a two-
input Type-3 ANFIS with 9 rules [7],[8],[26].
III. CONSEQUENT PARAMETERS
ESTIMATION
According to the proposed simple ANFIS
architecture (shown in Fig. 2), it can be observed
that given the values of the premise parameters, the
overall output can be expressed as a linear
combination of the consequent parameters. More
explicitly, the output Z in Fig.2 can be rewritten as:
222222111111
2211
)()()()()()( rwqywpxwrwqywpxw
fwfwZ

 (
2
)
By fixing the elements included in the normalized
firing strength of the rules and the above
expression in Equation 2 is linear in the consequent
parameters (p1, q1, r1, p2, q2, and r2) [13]. Therefore,
the matrix form of Equation 2 can be written as:
AX = B (3)
Where X is an unknown vector whose elements are
the linear consequent parameters of all rules. Given
that the number of these linear consequent
parameters is (M) and the number of training input
output data pairs is (N), then the dimensions of X,
A and B are N×M, M×1 and N×1, respectively.
Since N is always greater than M, the system of
equations in (3) is over-determined and generally
there is no exact or unique solution that can be
attained. Instead, a least-squares estimate (LSE) of
X is sought to minimize the squared error .
2
B-AX This is a standard problem that forms
the base in many applications like linear regression,
adaptive filtering and signal processing. The most
well known formula for solving such over-
determined system of equations uses the pseudo-
inverse of X:
X* = (AT
A)-1
AT
B (4)
Where AT is the transpose of A and (AT
A)-1
AT
is
the pseudo inverse of A if (AT
A ) is non-singular
[7], [20], [24], [25], [27], [28], [29], [30], [31].
IV. DATA
In this study three different data sets were used to
determine the best suitable subset numbers. These
254, 504 and 1006 points whose latitude, longitude,
ellipsoidal height and orthometric height are known
were used to construct Fuzzy models in the region.
The points are homogenously distributed and
randomly selected in Istanbul; the point density is
nearly one point in 20, 10 and 5 km2 respectively.
The data covers the region between 41° 30´2.78”
>> 40° 48´13”.75 and 29° 54´ 24”.24 >> 27° 59´
3”.05. The standard deviation of the ellipsoidal
heights after the adjustment of the network has
been found to be ± 2.56 cm [32]. To check for the
calculation, randomly selected 173 points which
had not been included in the preparation of the
Fuzzy models were used.
V. RESULTS
In this study, two different fuzzy models were
constructed. First group fuzzy model is constructed
with two inputs (latitude and longitude taken as
inputs) and one output (geoid height as taken
output) and the second group fuzzy model is
formed by three inputs (latitude, longitude and
ellipsoidal heights taken as inputs) and one output
(geoid height as taken output). Inputs were divided
into different subsets which give the best results in
both fuzzy models. Therefore, inputs are divided
into 9, 9 and 14 different subsets in two inputs and
one output fuzzy model using 254, 504 and 1006
points respectively. Likewise, inputs are divided
into 4, 4 and 5 different subsets in three inputs and
one output fuzzy model using 254, 504 and 1006
points respectively.
The efficiency of the ANFIS approximation was
compared to the results obtained by the
GPS/levelling method. The performance of the
ANFIS approximation was tested at the model
points and at the 173 test points. The differences
between geoid heights obtained by the
GPS/levelling method and the Fuzzy model results,
number of lineer and non-lineer parameters are
summarised in Table 1, Table 2 and Table 3.
y
x
1 4 7
2 5 8
3 6 9

It can be seen that the RMSE ( which is calculated
with  n
vv where v = error and n = number of
points) values vary from ± 2.21 cm to ± 5.45 cm
and from
Table 1: Summary of the resulsts obtained by 254 points
Number of
subsets
Number of
lineer
parameters
Number of
non- lineer
parameters
Total
parameters
Number of
rules
MSE at Model
Points
MSE at
TestPoints
2x2 12 8 20 4 5.45 6.33
3x3 27 12 39 9 4.47 5.84
4x4 48 16 64 16 4.04 5.32
5x5 75 20 95 25 3.59 4.58
6x6 108 24 132 36 3.18 4.67
7x7 143 28 171 49 2.96 5.92
8x8 192 32 224 64 2.74 7.63
9x9 243 36 279 81 2.60 16.61
10x10 300 40 340 100 2.21 53.22
2x2x2 32 12 44 8 3.80 5.56
3x3x3 108 18 126 27 2.71 4.93
4x4x4 256 24 280 64 1.64 12.18
5x5x5 500 30 530 125 1.09 222.33
Table 2: Summary of the resulsts obtained by 254 points
Number of
subsets
Number of
lineer
parameters
Number of
non- lineer
parameters
Total
parameters
Number of
rules
MSE at Model
Points
MSE at
TestPoints
2x2 12 8 20 4 5.37 5.99
3x3 27 12 39 9 4.43 4.97
4x4 48 16 64 16 4.04 4.87
5x5 75 20 95 25 3.72 4.70
6x6 108 24 132 36 3.36 4.25
7x7 143 28 171 49 3.14 4.07
8x8 192 32 224 64 2.91 3.85
9x9 243 36 279 81 2.82 3.51
10x10 300 40 340 100 2.42 4.68
2x2x2 32 12 44 8 3.93 4.79
3x3x3 108 18 126 27 3.09 4.51
4x4x4 256 24 280 64 2.46 3.88
5x5x5 500 30 530 125 1.90 5.56
Table 3: Summary of the results obtained by 254 points
Number of
subsets
Number of
lineer
parameters
Number of
non- lineer
parameters
Total
parameters
Number of
rules
MSE at Model
Points
MSE at
TestPoints
2x2 12 8 20 4 4.84 6.38
3x3 27 12 39 9 4.39 5.96
4x4 48 16 64 16 4.10 5.62
5x5 75 20 95 25 3.62 5.15
6x6 108 24 132 36 3.49 4.87
7x7 143 28 171 49 3.08 4.58
8x8 192 32 224 64 3.03 4.32
9x9 243 36 279 81 3.00 4.11
10x10 300 40 340 100 2.73 3.85
11x11 363 44 407 121 2.61 3.61

12x12 432 48 480 144 2.29 3.22
13x13 507 52 559 169 2.33 2.93
14x14 588 76 664 196 2.17 3.10
15x15 675 100 775 225 2.07 3.86
2x2x2 32 12 44 8 3.95 5.12
3x3x3 108 18 126 27 3.04 4.46
4x4x4 256 24 280 64 2.54 4.03
5x5x5 500 30 530 125 2.22 3.67
6x6x6 864 36 900 216 2.15 5.28
± 4.58 cm to ± 53.22 cm for model and test points
using 254 points by two inputs and one output
fuzzy models, respectively. On the other hand,
numbers of total parameter vary from 20 to 279 in
the models. Subsets divided by 5x5 can be used as
the best fuzzy model in this group. RMSE values
vary from ± 1.09 cm to ± 3.80 cm and from ± 4.93
cm to ± 222.33 cm for model and test points using
254 points by three inputs and one output fuzzy
models, respectively. On the other hand, numbers
of total parameter vary from 44 to 530 in the
models. Subsets divided by 3x3x3 can be used as
the best fuzzy model in this group
RMSE values also vary from ± 2.42 cm to ± 5.37
cm and from ± 3.51 cm to ± 5.99 cm for model and
test points using 504 points by two inputs and one
output fuzzy models, respectively. On the other
hand, numbers of total parameter again vary from
20 to 279 in the models. Subsets divided by 9x9
can be used as the best fuzzy model in this group
RMSE values vary from ± 1.90 cm to ± 3.93 cm
and from ± 3.88 cm to ± 5.56 cm for model and test
points using 504 points by three inputs and one
hand, numbers of total parameter also vary from 44
to 530 in the models. Subsets divided by 4x4x4 can
be used as the best fuzzy model in this group
RMSE values also vary from ± 2.07 cm to ± 4.84
cm and from ± 2.93cm to ± 6.38 cm for model and
test points using 1006 points by two inputs and one
hand, numbers of total parameter again vary from
20 to 775 in the models. Subsets divided by 13x13
can be used as the best fuzzy model in this group
RMSE values vary from ± 2.15 cm to ± 3.95 cm
and from ± 3.67 cm to ± 5.28 cm for model and test
points using 1006 points by three inputs and one
hand, numbers of total parameter also vary from 44
to 900 in the models. Subsets divided by 5x5x5 can
be used as the best fuzzy model in this group
According to this study, the best fuzzy models can
be constructed with total number of parameters is
nearly half of the number of points used in fuzzy
models. For example, a two inputs and one output
fuzzy model can be formed by dividing five subsets
of each inputs using 254 points. Number of lineer
parameters is 75 and number of non-lineer
parameters is 20 and total number of parameters is
95. This is nearly half of the total number of points
to contruct fuzzy model. A two inputs and one
output fuzzy model also can be formed by dividing
thirteen subsets of each inputs using 1006 points.
Number of lineer parameters is 507 and number of
non-lineer parameters is 52 and total number of
parameters is 559. This is again nearly half of the
total number of points to contruct fuzzy model.
In the same way, a there inputs and one output
fuzzy model can be constructed by dividing four
subsets of each inputs using 504 points. In this
case, Number of lineer parameters is 256 and
number of non-lineer parameters is 24 and total
number of parameters is 280. This is also nearly
half of the total number of points to contruct fuzzy
model.
The first paragraph under each heading or
subheading should be flush left, and subsequent
paragraphs should have a five-space indentation. A
colon is inserted before an equation is presented,
but there is no punctuation following the equation.
All equations are numbered and referred to in the
text solely by a number enclosed in a round bracket
(i.e., (3) reads as "equation 3"). Ensure that any
miscellaneous numbering system you use in your
paper cannot be confused with a reference [4] or an
equation (3) designation.
VI. CONCLUSION
Adaptive Network based Fuzzy Inference Systems
(ANFIS) can be used as a computation tool for
many areas such as geodetic sciences, civil
engineering, mechanical engineering, electrical

engineering and many other areas. Determination
of number of subsets is critical in ANFIS because it
cannot be determined easily. This is normally a
trial and error job. However, if how many subsets
can be selected known approximately, computation
time is reduced dramatically.
According to this study, the best fuzzy models can
be constructed with total number of parameters is
nearly half of the number of points used in fuzzy
models. To construct a two inputs and one output
fuzzy model using 254 points, inputs can be
divided five subsets to get best results. In the same
way, to form a two inputs and one output fuzzy
model using 1006 points, inputs can be divided
thirteen subsets to get best results. Again, dividing
five subsets gives the best results a three inputs and
one output fuzzy model using 1006 points.
Another factor effects result of ANFIS is type of
membership function. Studies show that Gauss and
Gauss2 membership functions give the best results.
While using ANFIS, one has to care about some
aspects for example, the number of parameters
(both premise and consequent) in ANFIS has to be
less than the number of training data pairs. This is
to avoid the over fitting phenomenon, which does
not allow a generalisation of the established fuzzy
inference system.
REFERENCES
[1]. Seager, J., Collier, P., Kirby, J., 1999. Modelling
geoid undulations with an artificial neural
network. IIEEE, International Joint Conference 5,
10-16 July 1999 Page(s):3332 – 3335.
[2]. Kotsakis, C. and Sideris, M. G., 1999. On the
adjustment of combined GPS/levelling/ geoid
networks. J Geod 73:412–421.
[3]. Engelis, T., Rapp, R. H., Bock, Y., 1985.
Measuring orthometric height differences with
GPS and gravity data.ManuscriptaGeodaetica 10
(3), 187-194.
[4]. Featherstone, W. E., 2001. Absolute and relative
testing of gravimetric geoid models using Global
Positioning System and orthometric height data.
Computers & Geosciences 27, 807-814.
[5]. Torge W, 2001. Geodesy. Walter de Gruyter,
Berlin, 416 pages.
[6]. Bajracharya, S., 2003, “Terrain effects on geoid
determination”, Ph.D. Thesis, University of
Calgary, Calgary, Alberta, Canada.
[7]. Akyılmaz, O., Ayan, T., Özlüdemir, T., 2003.
Geoid surface approximation by using Adaptive
Network Based Fuzzy Inference Systems,
AVN,110 (8-9) p. 308 – 315.
[8]. Jyh- Shing, R. J., 1993. ANFIS: adaptive –
network based fuzzy inference system. IEEE
Transactions on systems, Man, and Cybernetics,
23, (3), 665-685.
[9]. Cabalar, A.F., Cevik, A., Gokceoglu, C., 2012.
“Some applications of Adaptive Neuro- Fuzzy
Inference System (ANFIS) in geotechnical
engineering”, Computers and Geotechnics 40
(2012) 14–33.
[10]. Castellanos, F., James, N., 2009. “Average
Hourly Wind Speed Forecasting with ANFIS”,
11th American conference on Wind Engineering,
San Juan, Puerto Rico.
[11]. Kassem, Y., Çamur, H., bennur, K.E., 2018.
“Adaptive Neuro-Fuzzy Inference System
(ANFIS) and Artificial Neural Network (ANN)
for Predicting the Kinematic Viscosity and
Density of Biodiesel-Petroleum Diesel Blends”,
American Journal of Computer Science and
Technology, 2018; 1(1): 8-18.
[12]. Khademi, F., Jamal, S.M., Deshpande, N.,
Londhe, S., 2016. “Predicting strength of
recycled aggregate concrete using Artificial
Neural Network, Adaptive Neuro-Fuzzy
Inference System and Multiple Linear
Regression”, International Journal of Sustainable
Built Environment (2016) 5, 355–369.
[13]. Takagi, T., Sugeno, M., 1983. Derivation of
fuzzy control rules from human operator’s control
actions. Proc.Of the IFAC Symp.on Fuzzy
Information, Knowledge Representation and
Decision Analysis, 55-60, July, Marseilles,
France.
[14]. Reda-Taha, M. M.; Noureldin, A. and El-Sheimy,
N., 2003. Improving INS/GPS Positioning
Accuracy During GPS Outages Using Fuzzy
Logic, Proceedings of the 16th annual Technical
Meeting of Satellite Division, Institute of
Navigation (ION) GPS-GNSS, Portland, Oregon,
Sept., pp. 499-508.
[15]. Yılmaz, M., 2005. “Geoid research in Istanbul
metropolitan area”, Ph. D. Thesis, Institute of
Science and Technology, Istanbul Technical
University, Turkey (in Turkish).
[16]. Jyh- Shing, R. J., 1991. Fuzzy modelling using
generalized neural networks and kalman filter

algorithm. In Proc. of the Ninth National
Conference on Artificial Intelligence (AAAI-91),
762-767, July 15-19, Anaheim, California, USA.
[17]. Khan, Y., Chai, S.S., 2017. “Ensemble of ANN
and ANFIS for Water QualityPrediction and
Analysis - A Data Driven Approach”, Journal of
Telecommunication, Electronic and Computer
Engineering, Vol. 9 No. 2-9
[18]. Khoshnevisan, B., Rafiee, S., Omid, M.,
Mousazadeh, H., 2014. “Development of an
intelligent system based on ANFIS for predicting
wheat grain yield on the basis of energy inputs”,
Information Processing in Agriculture 1 (2014)
14–22.
[19]. Masrur Ahmed, A.A., Syed Mustakim, A.S.,
2015. “Application of adaptive neuro fuzzy
inference system (ANFIS) to estimate the
biochemical oxygen deman (BOD) of Surma
River” Journal of King Saud University –
Engineering Sciences (2017) 29, 237–243.
[20]. Mathur, N., Glesk, I., Buis, A., 2016.
“Comparison of adaptive neuro-fuzzy inference
system (ANFIS) and Gaussian processes for
machine learning (GPML) algorithms for the
prediction of skin temperature in lower limb
prostheses”, Medical Engineering and Physics 38
(2016) 1083–1089.
[21]. Patel, J., Parekh, F.,2014. “Forecasting Rainfall
Using Adaptive Neuro-Fuzzy Inference System
(ANFIS)”, International Journal of Application or
Innovation in Engineering & Management
(IJAIEM), Volume 3, Issue 6., 262-269.
[22]. Ramesh, K., Kesarkar, A. P., Bhate, J.,
VenkatRatnam, M. and Jayaraman, A., 2015.
“Adaptive neuro-fuzzy inference system for
temperature and humidity profile retrieval from
microwave radiometer observations”, Atmos.
Meas. Tech., 8, 369–384.
[23]. Sahin, M., and Erol, R., 2017. “A Comparative
Study of Neural Networks and ANFISfor
Forecasting Attendance Rate of Soccer Games”,
Math. Comput.Appl., 22, 43.
[24]. Takagi, T., Sugeno, M., 1985. Fuzzy
identification of systems and its application to
modelling and control,. IEEE Trans. on Systems,
Man and Cybernetics 15: 116-132.
[25]. Walid, A. H., 2005. “Accuracy Enhancement of
Integrated MEMS-IMU/GPS Systems for Land
Vehicular Navigation Applications”, Ph. D.
Thesis, Department of Geomatics Engıneering,
Unıversity of Calgary, Calgary, Alberta, Canada.
[26]. Jyh- Shing, R. J., 1995. Neuro – fuzzy modelling
and control. Proceedings of the IEEE, 83, No:3,
378-406.
[27]. Akyılmaz, O., 2005. “Applications of Soft
Computing methods in Geodesy”, Ph. D. Thesis,
Institute of Science and Technology, Istanbul
Technical University, Turkey (in Turkish).
[28]. Astrom, K. J. and Wittenmark, B., 1984.
Computer Controller Systems: Theory and
Design. Prentice-Hall, Inc., Englewood Cliffs,
New Jersey, 1984.
[29]. Goodwin, G. C. and Sin, K. S., 1984. Adaptive
filtering prediction and control. Prentice - Hall,
Englewood Cliffs, N. J., 540 pages.
[30]. Ljung, L., 1987. System identification: theory for
the user. Prentice-Hall, Englewood Cliffs, N. J.,
384 pages
[31]. Strobach, P., 1990. Linear prediction theory: a
mathematical basis for adaptive systems. Springer
–Verlag New York, 418 pages.
[32]. Ayan T, Deniz R, Arslan E, Çelik RN, Denli HH,
Akyılmaz O, Ozsamli C, Özlüdemir MT, Erol S,
Erol B, Acar M, Mercan H, Tekdal E (2006).
Istanbul GPS network technical report, ITU Civil
Engineering Faculty, Department of Geodesy and
Photogrammetry Engineering, Istanbul (in
Turkish).

How to Decide the Best Fuzzy Model in ANFIS

More Related Content

What's hot (20)

Similar to How to Decide the Best Fuzzy Model in ANFIS (20)

More from International Journal of Modern Research in Engineering and Technology (20)

Recently uploaded (20)

How to Decide the Best Fuzzy Model in ANFIS