(IJCSIS) International Journal of Computer Science and Information Security,Vol. 8, No. 8, 2010
.
An Improved Fuzzy Time Series Model For Forecasting
Ashraf K. Abd-Elaal
1
Department of Computer and Information Sciences
The High Institute of Computer ScienceSohag, Egyptashrafsohag@yahoo.com
Hesham A. Hefny
Department of Computer and InformationSciences,
Institute of Statistical Studies and Research,Cairo University, Egypthehefny@ieee.org
Ashraf H. Abd-Elwahab
Department of Computer Sciences
Electronics Research Institute National Center for ResearchCairo, Egypt
Abstract
— Researchers introduce in this paper, an efficient fuzzytime series forecasting model based on fuzzy clustering to handleforecasting problems and improving forecasting accuracy. Eachvalue (observation) is represented by a fuzzy set. The transitionbetween consecutive values is taken into account in order tomodel the time series data. Proposed model employed eight mainsteps in time-invariant fuzzy time-series and time-variant fuzzytime series models to increase the performance of the proposedfuzzy time series model. The method of FCMI is integrated in theprocesses of fuzzy time series to partition datasets. The proposedmodel has been implemented to forecast the world production of iron and steel and the enrollments of the University of Alabama.The proposed model provide higher accuracy in forecasting. Ourresults show that this approach can lead to satisfactoryperformance for fuzzy time series
Keywords- forecasting; fuzzy Clustering; fuzzy time series; iron.
I.
I
NTRODUCTION
Traditional forecasting methods can deal with manyforecasting cases, but they cannot solve forecasting problems inwhich the historical data are linguistic values. Song andChissom [12] presented the concept of fuzzy time series basedon the historical enrollments of the University of Alabama.They presented the time-invariant fuzzy time series model andthe time-variant fuzzy time series model based on the fuzzy settheory for forecasting the enrollments of the University of Alabama.The fuzzy forecasting methods can forecast the data withlinguistic values. Fuzzy time series do not need to turn a non-stationary series into a stationary series and do not require morehistorical data along with some assumptions like normality postulates. Although fuzzy forecasting methods are suitable for incomplete data situations, their performance is not alwayssatisfactory [9,11].Huarng [6] proposed heuristic models; by integrating problem-specific heuristic knowledge to improve forecasting.
1
Corresponding Author:
Ashraf K. Abd-Elaal
Tsaur, et al [14] proposed an analytical approach to findthe steady state of fuzzy relation matrix to revise the logicforecasting process. Based on the concept of fuzziness inInformation Theory, the concept of entropy is applied tomeasure the degrees of fuzziness when a time-invariant relationmatrix is derived. In order to show the forecasting performance, the best fitted regression equations are applied tocompare with the proposed method.Yu [15] proposed weighted models to tackle two issues infuzzy time series forecasting; namely, recurrence andweighting. Weighted fuzzy time series models appear quitesimilar to the weight functions in local regression models;however, both are different. The local regression models focuson fitting using a small portion of the data, while the fuzzyrelationships in weighted fuzzy time series models areestablished using the possible data from the whole of thedatabase.Jilani and Burney [7] presented two new multivariatefuzzy time series forecasting methods. These methods assumem-factors with one main factor of interest. Stochastic fuzzydependence of order k is assumed to define general methods of multivariate fuzzy time series forecasting and control.Cheng et al [4] proposed a novel multiple-attribute fuzzytime series method based on fuzzy clustering. The methods of fuzzy clustering were integrated in the processes of fuzzy timeseries to partition datasets objectively and enable processing of multiple attributes.Abd Elaal et al [1-2] proposed a novel forecasting fuzzytime series model depend on fuzzy clustering for improvingforecasting accuracy. Kai et al [8] proposed a novel forecastingmodel for fuzzy time series using K-means clusteringalgorithm for forecasting.In this paper, researchers propose an efficient fuzzy timeseries forecasting model based on fuzzy clustering to handleforecasting problems and improving forecasting accuracy. Eachvalue (observation) is represented by a fuzzy set. The transition between consecutive values is taken into account in order tomodel the time series data.
11http://sites.google.com/site/ijcsis/ISSN 1947-5500
(IJCSIS) International Journal of Computer Science and Information Security,Vol. 8, No. 8, 2010
.
II.
R
ELATED WORKS
In this section, two related works including: fuzzy clusteringand fuzzy time series.
A.
Fuzzy clustering (FCMI)
Fuzzy c-means (FCM) is a method of clustering whichallows one piece of data to belong to two or more clusters.Fuzzy C-Mean Iterative assume that: the existence of patternspace X={x
1
, x
2
,…, x
m
) and c fuzzy clusters, whose centershave initial values y
10
, y
20
,…,y
c0
. Every iteration themembership function values updated and the cluster centersalso. The process terminates when the difference between twoconsecutive clusters centers do not exceed a given tolerance[5].
)()(
k i jk ij
y xd
−=
(1)Fuzzy clustering is carried out through an iterative optimizationof the objective function
ij
d
, with the update of membership
ij
u
and the cluster centers
i
y
by:
11)1/(2
)()()(
−=−
=
∑
cl k ljk ijk ij
d d u
β
(2)
∑∑
==+
=
m jk ijm j jk ijk i
u xu y
1)(1)()1(
(3)This iteration will stop when
ε
<
−
∑
=+
2/112)()1(
cik ik i
y y
(4)
B.
Fuzzy time series
Song and Chissom [13] presented the concept of fuzzytime series based on the historical enrollments of the Universityof Alabama. Fuzzy time series used to handle forecasting problems. They presented the time-invariant fuzzy time seriesmodel and the time-variant fuzzy time series model based onthe fuzzy set theory for forecasting the enrollments of theUniversity of Alabama. The definitions and processes of thefuzzy time-series presented by Song and Chissom are describedas follows [6,12].
Definition 1
. (FTS) Assume Y (t) (t = . . 0, 1, 2, . . .) is a subsetof a real numbers. Let Y (t) be the universe of discoursedefined by the fuzzy set f
i
(t). If F(t) is a collection of f
1
(t), f
2
(t).. . then F(t) is defined as a fuzzy time-series on Y (t) (t = . . . ,0, 1, 2, . . .).
Definition 2
. (FTSRs) If there exists a fuzzy logicalrelationship R(t − 1, t), such that F(t) = F(t − 1) × R(t − 1, t),where “×” represents an operation, then F(t) is said to beinduced by F(t − 1). The logical relationship between F(t) andF(t − 1) is F(t − 1)
F(t).
Definition 3
. (FLR) suppose F(t − 1) = A
i
and F(t) = A
j
. Therelationship between two consecutive observations, F(t) and F(t− 1), referred to as a fuzzy logical relationship, can be denoted by A
i
A
j
, where Ai is called the Left-Hand Side (LHS) andA
j
the Right-Hand Side (RHS) of the FLR.
Definition 4.
(FLRG) All fuzzy logical relationships in thetraining dataset can be grouped together into different fuzzylogical relationship groups according to the same Left-HandSides of the fuzzy logical relationship. For example, there aretwo fuzzy logical relationships with the same Left-Hand Side(A
i
): A
i
A
j1
and A
i
A
j2
. These two fuzzy logicalrelationships can be grouped into a fuzzy logical relationshipgroup A
i
A
j1
A
j2
.
Definition 5.
(IFTS & VFTS) Assume that F(t) is a fuzzy time-series and F(t) is caused by F(t − 1) only, and F(t) = F(t − 1) ×R(t − 1, t). For any t, if R(t − 1, t) is independent of t, then F(t)is named a time-invariant fuzzy time-series, otherwise a time-variant fuzzy time-series.
a)
Song and Chissom model
Song and Chissom employed five main steps in time-invariant fuzzy time-series and time-variant fuzzy time seriesmodels as follows:
Step 1
: Define the universe of discourse U. Define the universeof discourse for the observations. According to the issuedomain, the universe of discourse for observations is definedas,U=[D
min
– D
1
, D
max
+ D
2
] (5)where, D
min
is the minimum value,D
max
is the maximum value,D
1
, D
2
is the positive real numbers.
Step 2
: Partition universal of discourse U into equal intervals.
Step 3
: Define the linguistic terms. Each linguistic observation,A
k
can be defined by the intervals u
1
,u
2
,...,u
n
, as follows:
12k1kk1n1n
10.5k1uu0.50.51k uuu0.51knuu
A 2 kn1
− +−
+ =+ =
= + + ≤ ≤ −
(6)
12http://sites.google.com/site/ijcsis/ISSN 1947-5500
(IJCSIS) International Journal of Computer Science and Information Security,Vol. 8, No. 8, 2010
.
Step 4
: Fuzzify the historical data. Each historical data can befuzzified into a fuzzy set.
Step 5
: Build fuzzy logic relationships. Build fuzzy logicrelationships. Two consecutive fuzzy sets A
i
(t-1)and A
j
(t) can be established into a single FLR as A
i
A
j
.III.
P
ROPOSED MODEL
In this section we proposed an efficient fuzzy time seriesforecasting model based on fuzzy clustering to handleforecasting problems and improving forecasting accuracy.Most researchers have been taken the same way according to processes of the fuzzy time-series, which are presented by Songand Chissom, but we introduce a novel model based on fuzzyclustering to determine the membership values not as Song andChissom model, and to increase the performance. Proposedmodel employed eight main steps in time-invariant fuzzy time-series and time-variant fuzzy time series models as follows:
Step 1: Cluster data into c clusters:
Apply fuzzy clusteringon a time series Y(t) with n observation to cluster this timeseries into c (2 ≤ c ≤ n) clusters. FCMI is used because it is themost popular one and well known in fuzzy clustering field.
Step 2: Determine membership values for each cluster:
Inthis step, membership values is determining after doing fuzzycluster. The proposed model selected the maximummembership grade of each value for each cluster which it belong to.
Step 3: Rank each cluster:
Proposed model ranking clusters by the center of each cluster, where first cluster has theminimum center, and last cluster has the maximum center.
Step 4: Define the universe of discourse U:
In this step, the proposed model defines the universe of discourse as Song andChissom were defined it as in (5).
Step 5: Partition universal of discourse U into equalintervals:
According to this step, the proposed model, partitionthe universe of discourse into c intervals.
Step 6: Fuzzify the historical data:
In this step, proposedmodel fuzzufy historical data, where the proposed modeldetermine the best fuzzy cluster to each actual data
Step 7: Build fuzzy logic relationships:
Proposed model inthis step build fuzzy logic relationship as definition 3. if F(t−1)= Ai and F(t) = Aj then the relationship between twoconsecutive observations: Ai
Aj
Step 8: Calculate forecasting outputs:
The forecasting valuefor each cluster is calculated by proposed model as:
∑
=+++=
m j jdf mdf df df i A forecaste
1mXx...1Xx11Xx1)(
(7)Where dfj is the membership grade,Xj is the actual value.
A.
Evaluating of the proposed model
To evaluating the performance of the proposed model, theresearchers compare the forecasting values of enrollments of the University of Alabama with some famous models such asJilani and Burney [7], Tsaur and Yang [14], Yu [15], Kai et al[8], and Cheng, et al [4].The forecasting accuracy is compared by using (NRMSE) Normalized Root Mean Square Error. NRMSE, in statistic isthe square root of the sum of the squared deviations betweenactual and predicted values divided by the sum of the square of actual values.
∑∑
==
−=
N ii N iii
actual predict actual NRMSE
1212
)()(
(8)In this study, to evaluate the forecasting accuracy of the proposed model, the researchers use the enrollments of theUniversity of Alabama as the forecasting target in the existingforecasting models.Based on the enrollments of the University of Alabamafrom 1971 to 1992, we can get the universe of discourseU=[13055,19337], partition U into 7 equal intervals, D
1
=13,and D
2
=55. Hence, the intervals are u
1
; u
2
; u
3
; u
4
; u
5
; u
6
; u
7
;where :-u
1
=[13024.00, 13933.71]u
2
=[13933.71, 14843.43],u
3
=[14843.43, 15753.14],u
4
=[15753.14, 16662.86],u
5
=[16662.86, 17572.57],u
6
=[17572.57, 18482.29],u
7
=[18482.29, 19392.00],Table I lists the enrollment of the University of Alabamafrom 1971 to 1992, and membership grades of enrollments for each linguistic. Define the fuzzy set A
i
using the linguisticvariable "Enrollments of the University of Alabama", let A
1
=(very very few), A
2
= (very few), A
3
= (few), A
4
= (moderate),A
5
= (many), A
6
= (many many), A
7
= (too many).The proposed model selected the maximum membership grade for each cluster, the forecasting value for each cluster calculatingas in (7):
135631(1972)x1)1(
==
A forecaste
151458.0(1984)x8.0
)2(
==
A forecaste
13http://sites.google.com/site/ijcsis/ISSN 1947-5500
Search History:
Result 00 of 00
00 results for result for
p.
An Improved Fuzzy Time Series Model For Forecasting
Researchers introduce in this paper, an efficient fuzzy time series forecasting model based on fuzzy clustering to handle forecasting problems and improving forecasting accuracy. Each value (observation) is represented by a fuzzy set. The transition between…
(More)
545
Reads
12
Readcasts
15
Embed Views
Recommended
More From This User
Notes
Load more
