Hybrid Personalized Recommender System Using Modified Fuzzy C-Means Clustering Algorithm

S. K. Shinde & U. V. Kulkarni
International Journal of Artificial Intelligence And Expert Systems (IJAE), Volume (1): Issue (4) 88
Hybrid Personalized Recommender System Using Modified
Fuzzy C-Means Clustering Algorithm
Subhash K. Shinde skshinde@rediffmail.com
Department of Computer Engineering,
Bharati Vidyapeeth College of Engineering,
Navi Mumbai 400 614, India.
Uday V. Kulkarni kulkarniuv@yahoo.com
Department of Computer Science and Engineering,
SGGS Institute of Engineering and Technology,
Nanded 431605, India
Abstract
Recommender Systems apply machine learning and data mining techniques for
filtering unseen information and can predict whether a user would like a given
resource. This paper proposes a novel Modified Fuzzy C-means (MFCM)
clustering algorithm which is used for Hybrid Personalized Recommender
System (MFCMHPRS). The proposed system works in two phases. In the first
phase, opinions from the users are collected in the form of user-item rating
matrix. They are clustered offline using MFCM into predetermined number
clusters and stored in a database for future recommendation. In the second
phase, the recommendations are generated online for active users using
similarity measures by choosing the clusters with good quality rating. We
propose coefficient parameter for similarity computation when weighting of the
users’ similarity. This helps to get further effectiveness and quality of
recommendations for the active users. The experimental results using Iris
dataset show that the proposed MFCM performs better than Fuzzy C-means
(FCM) algorithm. The performance of MFCMHPRS is evaluated using Jester
database available on website of California University, Berkeley and compared
with fuzzy recommender system (FRS). The results obtained empirically
demonstrate that the proposed MFCMHPRS performs superiorly.
Keywords: Fuzzy C-means, Modified Fuzzy C-means, Personalized Recommender System.
1. INTRODUCTION
Modern consumers are inundated with choices. Electronic retailers and content providers offer a
huge selection of products with unprecedented opportunities to meet a variety of special needs
and tastes. Matching consumers with the most appropriate products is the key to enhancing user
satisfaction and loyalty. Therefore, more retailers have become interested in recommender
systems, which analyze patterns of user interest in products to provide personalized
recommendations that suit a user’s taste. As good personalized recommendations can add
another dimension to the user experience, e-commerce leaders like Amazon.com and Netflix
have made recommender systems a salient part of their websites [1]. Such systems are

particularly useful for entertainment products such as movies, music, jokes, and TV shows. Many
customers will view the same movie, and each customer is likely to view numerous different
movies. Customers have proven willing to indicate their level of satisfaction with particular
movies, so a huge volume of data is available about which movies appeal to which customers.
Companies can analyze this data to recommend movies to particular customers.
The remainder of this paper is organized as follows. The section 2 summarizes the different
strategies for recommender systems and their drawbacks. The proposed clustering based hybrid
personalized recommender system is described in the section 3. The section 4 illustrates
experimental setup of the proposed recommendation system. This section also gives
performance evaluation with the existing algorithms. Finally, the section 5 concludes the paper.
2. RECOMMENDER SYSTEM STRATEGIES
In the recent years web personalization has undergone through tremendous changes. The
content [2, 3], collaborative [4, 5] and hybrid [6] based filtering are three basic approaches used
to design recommendation systems.
The content based filtering [7] relies on the content of an item that user has experienced before.
The content based information filtering has proven to be effective in locating text, items that are
relevant to the topic using techniques such as Boolean queries, vector space queries etc.
However, content based filtering has some limitations. It is difficult to provide appropriate
recommendation because all the information is selected and recommended based on the content.
Moreover, the content based filtering leads to overspecialization i.e. it recommends all the related
items instead of the particular item liked by the user.
The collaborative-filtering [8] aims to identify users who have relevant interests and preferences
by calculating similarities and dissimilarities between their profiles. The idea behind this method is
that to one’s search the information collected by consulting the behavior of other users who
shares similar interests and whose opinions can be trusted may be beneficial. The different
techniques have been proposed for collaborative recommendation; such as correlation based
method, semantic indexing etc. The collaborative filtering overcomes some of the limitations of
the content based filtering. The system can suggest items to the user, based on the rating of
items, instead of the content of the items which can improve the quality of recommendations.
However, collaborative filtering has some drawbacks. The first drawback is that the coverage of
rating could be very sparse thereby resulting in poor quality recommendation. In the case of the
addition of new items into database, the system would not be able to recommend until that item is
served to a substantial number of users known as cold-start. Secondly, when new users are
added, the system must learn the user preferences from the rating of users, in order to make
accurate recommendations. Moreover, these recommendation algorithms seem to be very
extensive and grow non-linearly when the number of users and items in a database increase. The
hybrid recommendation systems [9, 10, 11] combine content and collaborative based filtering to
overcome these limitations. As stated below, there are different ways of combining content and
collaborative based filtering [12].
i. Implementing these approaches separately and combining them for prediction.
ii. Incorporating some content based characteristics into collaborative approach and vice
versa.
iii. Constructing a general unified model that incorporates both content and collaborative
based characteristics.
The hybrid approach proposed in this paper extracts user’s current browsing patterns using web
usage mining, and forms a cluster of items with similar psychology to obtain implicit users rating
for the recommended item.
3. PROPOSED MFCMHPRS

We have developed and tested the MFCMHPRS for Jester dataset available on website of
California University, Berkeley. The system architecture has been partitioned into two main
phases; offline and online. The Fig. 1 depicts the architecture of MFCMHPRS with its essential
components.
The phase I is offline. It does the preprocessing and clustering. In this phase background data in
the form of user-item rating matrix is collected and clustered using the proposed approach which
is described in section 3.1.2. Once the clusters are obtained the cluster data along with their
centroids are stored for future recommendations. The phase II is online in which the
recommendation takes place for the active user. Here, similarity between active users and
clusters are calculated for choosing best clusters for making recommendations. The rating quality
of each item unrated by active user is computed in the chosen clusters. To generate the
recommendations, clusters are further selected based on rating quality of an item. The
recommendations are then made by computing the weighted average of the rating of items in the
selected clusters. The working of MFCMHPRS is described below in detail with the Jester
dataset.
Fig.1. System architecture of CBBCHPRS
Figure 1: The architecture of MFCMHPRS
Preprocessing phase
3.1.1 Normalization of data
User-item rating taken from Jester dataset rated in the scale of -10 to +10 is normalized in the
scale of 0 to 1, where 0 indicates that item is not rated by corresponding user. To facilitate the
discussion, running example shown in the Table 1 is used, where U1-U10 are the users and J1-J10
are the items (jokes) rated or unrated by users. The last row of Table 2 gives ratings of the active
user (U1
).

Users J1 J2 J3 J4 J5 J6 J7 J8 J9 J10
U1 0.15 0.94 0.06 0.13 0.16 0.11 0.05 0.72 0.09 0.29
U2 0.71 0.51 0.82 0.73 0.41 0.06 0.48 0.26 0.94 0.96
U3 0.00 0.00 0.00 0.00 0.95 0.96 0.95 0.96 0.00 0.00
U4 0.00 0.92 0.00 0.00 0.60 0.91 0.38 0.81 0.00 0.61
U5 0.92 0.74 0.32 0.26 0.58 0.60 0.85 0.74 0.50 0.79
U6 0.23 0.35 0.54 0.11 0.18 0.31 0.11 0.48 0.20 0.43
U7 0.00 0.00 0.00 0.00 0.93 0.05 0.89 0.94 0.00 0.00
U8 0.84 0.67 0.96 0.22 0.13 0.44 0.96 0.59 0.27 0.31
U9 0.34 0.35 0.07 0.19 0.10 0.51 0.27 0.09 0.14 0.44
U10 0.66 0.76 0.76 0.66 0.82 0.76 0.94 0.64 0.66 0.91
U1
0.38 0.71 0.00 0.00 0.20 0.00 0.64 0.27 0.00 0.59
TABLE 1: Running example of rating matrix from Jester data set after normalization in the range of
0 to 1
3.1.2 Modified Fuzzy C-means Clustering
Fuzzy C-Means algorithm also known as Fuzzy ISODATA, was introduced by Bezdeck [13] as an
extension to Dunn’s algorithm [14]. The FCM- based is the most widely used fuzzy clustering
algorithms in practice. However in FCM there are several constraints that affect the performance.
The first limitation is the selection of random centorids at initial level. So the algorithm takes more
time to find clusters. The second constraint is its inability to calculate the membership value if the
distances of data point is zero. Whereas, the proposed MFCM algorithm initially calculates
centorids appropriately and proposes new member function to calculate the membership value
even if the distances of data point is zero.
Let { }nx,.......,x,xX 21= where
n
ix ℜ∈ present a given set of feature data. The
objective of MFCM algorithm is to minimize the cost function formulated as
(1)
{ }cv,....,v,vV 21= are the cluster centers. The cluster centers are initially calculated as follows.
To determine the centroid of the cluster, all the patterns are applied to each of the pattern and the
patterns having Euclidian distance less than or equal to α (user defined value) are counted for all
the patterns. Later the pattern with the maximum count is selected as the centroid of the cluster.
If then 1+= ii DD for .p,i ,2,1 L= (2)
If maxD is the maximum value in the row vector D and indD is the index of maximum value
For instance the most appropriate centorids at the initial level are using centering process to form
three clusters of running example shown in the Table 1 are { }761 U,U,UV = .
( ) ( )
2
1 1
∑∑= =
−=
C
j
C
i
ji
m
ij vxV,UJ µ










≤−
=
αji RR
P
j 1
[ ] [ ]DmaxDD indmax = indRC =1

( ) CNijU ×
= µ is fuzzy partition matrix, in which each member ijµ indicates the degree
membership between the data vector ix and cluster .j The values of matrix U should satisfy the
following conditions
[ ]10,∈µ , N,i L1=∀ , C,j L1=∀ (3)
∑=
=
C
j
ji
1
1µ , N,i L1=∀ (4)
Appropriate initialize of the membership matrix U using
( ) ().fr,v,xf −= 1 , (5)
where ()





<<
=
≥
=
parameter)sensitiveis(10if
0if0
1if1
γγ rr
r
r
.f ,
ji vxr −= , 1≥r , and if 1>r then 1.set toisγr
To satisfy the condition 2 and 3, divide the total sum of attributes to the each attribute for every
pattern.
The exponent [ ]∞∈ ,m 1 is the weighting exponent which determines the fuzziness of the
clusters. Minimization of the cost function [ ]V,UJ is nonlinear optimization problem, which can
be minimized with following iterative algorithm:
Step 1: Find appropriate centriods using equation (2).
Choose appropriate exponent m and termination criteria.
Step 2: Initialize the membership matrix U using equation (4)
Step 3: Calculate the cluster center V according to the equation:
( )
( )∑
∑
=
=
= N
i
m
ij
N
i
ij
m
ij
j
x
v
1
1
µ
µ
, C,j L1=∀ (6)
Step 4: Calculate new distances norm:
ji vxr −= , N,i L1=∀ , C,j L1=∀
Step 5: Update the fuzzy partition matrixU :
If 0>r (indicating that ( )ji vx ≠ )

∑=
−






=
C
k
m
ik
ij
ij
r
r
1
1
2
1
µ (7)
Else
1=jiµ
Step 6: If the termination criteria has been met, stop.
Else go to step 2.
A suitable termination criterion could be to evaluate the cost function (Eq. 1) and to see whether it
is below a certain tolerance value or if its improvement compared to the previous iteration is
below a certain threshold. Also the maximum number of iteration cycles can be used as a
termination criterion.
3.1.3 Computing centorid of each cluster
The proposed MFCM is used for clustering of the Jester data set. The clustering is resulted in the
three clusters with 90.=α and 010.=ε (ε termination criteria). The details of the clusters are
created and users in each cluster are shown in the Table 3. After clustering as stated in the
MFCM algorithm, knowing the members of each group, we have recomputed new centroids of
each cluster. As an example the cluster 3 has two members. Thus the centorid is the average of
all corresponding coordinates of the two members
C3= {(0.00+0.00)/2, (0.00+0.00)/2, (0.00+0.00)/2, (0.00+0.00)/2, (0.60+0.93/2), (0.91+0.05/2),
(0.00+0.00)/2, (0.00+0.00)/2, (0.00+0.00)/2, (0.00+0.00/2)}. Similarly, we have calculated the
centroids of the cluster 1 and 2.
Cluster No. Users Centroid
1 U2, U6, U8, U10 C1
2 U1, U4, U5, U9 C2
3 U3, U7 C3
TABLE 2: Users in each cluster with the centroid
Recommendation phase
This phase consists of two steps: (i) find the nearest neighbors and (ii) produce recommendations
set.
Find the nearest neighbors:
In order to find the nearest neighbors of the active user, it must measure the similarity of the
users. Calculate similarity between clusters centroids and active users. Select cluster that have
the highest similarity. We use cosine similarity algorithm to measure the similarity between active
user
1
u and cluster uc . User rating can be treated as a vector on an n-dimensional item space.
Assuming the rating of the n-dimensional item space rated by U1
and Cu is respectively vector
1
u and uc .

( ) ( )
u
u
uu
cu
cu
c,ucosc,usim
⋅
⋅
==
1
1
11
In most cases, the number of items usually one or two jointly rated by two users is few. Even the
rating of these items rated by the two users has high similarity. According to common sense we
can not judge the two users are similar; but the semblance of the two users is very high if we use
traditional similarity measurement method. In order to solve this problem, we introduce a
coefficient: the coefficient is large if there are many items that the two users jointly rate; on the
contrary, the coefficient is small. We suppose that the coefficient isk given as
( )
( )u
u
c,u
c,u
k 1
1
U
I
= .
( )uc,u1
I is represents the number of items in the intersection set that rated both by user
1
u
and uc , ( )uc,u1
U represents the number of items in the union set that rated both by user
1
u
and uc . The range of k is in between 0 and 1.
Hence the similarity between active users and cluster centriods computed as:
( ) ( ) k
cu
cu
c,ucosc,usim
u
u
uu ∗
⋅
⋅
==
1
1
11
(8)
For running example, the similarity value of active user of three clusters is shown in Table 3.
Choose the clusters having high similarity value.
Cluster1 Cluster 2 Cluster 3
( )uc,usim 1
0.53 1.13 1.25
k 0.25 0.39 0.80
( )uc,usim 1
* k 0.1325 0.4407 1.00
TABLE 3: Nearest neighbors of the active user
Produce recommendation data set
The predication rating of item i by
1
u is ( )iPu1 which is gained by the rating of nearest neighbors
set uc rated by active user
1
u , the computation method is as the follows:
( )
( ) ( ) ( )( )
( )( )∑
∑
∈
−∗
+=
u
u
c,u
u
cu
u
uu
uu
|c,usim|
iRiRc,usim
RiP
1
1
1
11
1
1
(9)

where 1
u
R - average rating of items rated by active user
1
u ;
( )uc,usim 1
- Similarity between active user and users clusters;
( )iRu - average rating of item i rated by all user;
( )iRu1 - rating of active user of item i .
According to the rating of items, we select N items (N is user define parameter) that have the
highest rating to compose recommendation set and recommend them to active users. The
predication rating of active user for running example is shown in Table 4.
Jokes Predicating Rating
Active User
( )1
u
J3 0.73
J4 0.43
J6 0.55
J9 0.42
TABLE 4: Predication rating for active user
Once the quality rating of each item is calculated, the recommendation to the active user is
provided, e.g., joke 3 predication rating up to 0.73 will be recommended and so on.
4. EXPERIMENTS
We have conducted a set of experiments to examine the effectiveness of our proposed
recommender system in terms of accuracy of neighbor-selection, cold start and recommendation
quality. In particular, we addressed the following issues [15, 16, 17, 18, 19].
i. How does the confidence parameter affect the performance of the prediction? In this paper,
we have conducted few experiments to show the accuracy of the prediction for different
settings of the parameter values.
ii. How does the neighbor-selection method affect the efficiency of prediction? Experiments are
conducted to examine the accuracy of MFCM algorithm for neighbor-selection.
iii. How do the clusters formed influence the prediction accuracy? Experiments are conducted to
examine the impact of clustering methods on the final performance of item or user content
based collaborative filtering.
iv. The performance MFCMHPRS is evaluated and compared with FRS using Precision, Recall.
The proposed MFCMHPRS is implemented in MATLAB version 7.2. The experiments are
conducted on a 2.0 GHz, Intel Pentium 4 PC with 512 MB memory, running Microsoft Windows
XP Professional.
4.1 Simulation results and performance evaluation
4.1.2 Performance evaluation of clustering
In order to check the performance of the proposed clustering algorithm, we have first applied the
algorithm to real data set, ‘Iris’ data, whose true classes are known. The Iris data set is available
in UCI repository (ftp://www.ics.uci.edu/ pub/machinelearningdatabases/), which includes 150
objects (50 in each of three classes – ‘Setosa’, ‘Versicolor’, and ‘Virginica’) having four variables
(‘sepal length’, ‘sepal width’, ‘petal length’, and ‘petal width’).

The performance was measured by the accuracy, which is the proportion of objects that are
correctly grouped together against the true classes. To investigate the performance more
objectively, a simulation study was carried out by generating artificial data sets repetitively and
calculating the average performance of the method.
We have applied the proposed MFCM, and FCM to create three clusters using this data
without the class information. The table 5 shows the result obtained using existing and proposed
clustering method.
Algorithms Setosa Versicolor Virginica
Computational
Time
FCM 50 34 66
13.5790
seconds
MFCM 50 39 61
11.3790
seconds
TABLE 5: Cluster result of Iris data by the proposed and traditional methods
The table 7 shows that the proposed MFCM clustering algorithm works superior than the
traditional algorithms because the algorithm calculates centroids and Initialize the membership
matrix properly instead of selecting randomly.
4.1.2 Performance evaluation of recommender system
The Jester dataset is available online on the site www.ieor.berkeley.edu/~goldberg/jester-data.
The Jester is a WWW Based Joke Recommender System, developed by University of California,
Berkeley. This data has 73421 user entered numeric rating for 100 jokes, ranging on real value
scale from -10 to 10. The experiments are performed on the small Jester dataset consisting of
user-item rating matrix of size 100 (users) ×10 (jokes) as shown in the Table 1.
The measurement method of evaluating the recommendation quality of recommendation system
mainly includes statistical precision measurement method and decision supporting precision
measurement method [20, 21]. Statistical precision measurement method adopts MAE (Mean
Absolute Error) to measure the recommendation quality [22]. MAE is a commonly used
recommendation quality measurement method. So we use MAE as the measurement criteria.
MAE calculates the irrelevance between the recommendation value predicted by the system and
the actual evaluation value rated by the user. We represent each pair of interest predicted rank as
<pi, qi>, pi is the system predicted value, qi is the user evaluation value. Basing on the entire <pi,
qi> pairs, MAE calculates the absolute error value |pi-qi| and the sum of all the absolute error
value, and then calculates their average value. If the MAE value is small, it indicates good
recommendation quality.
The predicted user rating set can be represented as{ }Np,,p,p L21 , its corresponding actual
user rating set can be represented as{ }Nq,,q,q L21 , the MAE can be defined as the
following[23]:
N
qp
MAE
N
i
ii∑=
−
= 1
(10)

In addition to MFCMHPRS, FRS [24] is also implemented to compare the performance with our
proposed system. Let us examine the influence of various nearest neighbor set on predictive
validity. We gradually increase the number of neighbors; the experiment result is shown in Table
6:
Size of Neighbor
Set
MAE
FRS MFCMHPRS
04 1.3272 1.2213
08 1.2531 1.2163
12 1.2615 1.2182
16 1.2480 1.2203
20 1.2573 1.2232
Table 6: Influence of various size of nearest neighbor set on predictive validity
As Fig. 2 shown, the MFCMHPRS has smaller MAE value than FRS in most cases, which means
that the sparseness has the less impact on our proposed algorithm.
4 6 8 10 12 14 16 18 20
1.2
1.22
1.24
1.26
1.28
1.3
1.32
1.34
MAE
Size of Neighbour set
FRS
MFCMHPRS
Figure: 2 MAE on each algorithm. (A small value means a better performance)
5. CONCULSIONS
This paper describes a novel fuzzy personalized recommender system that utilizes clustering of
user-item rating matrix through proposed MFCM and provides the recommendations for the
active user with good quality rating using similarity measures. The results from various
simulations using Iris data set shows that the proposed MFCM clustering algorithm performs
better than FCM clustering, which helps to improve the quality of rating. Through the experiment
analysis, it is found that the proposed MFCMHPRS performs better than FRS and the sparseness
has less impact on the proposed system.

REFERENCES
1. G. Adomavicius and A. Tuzhilin. “Toward the Next Generation of Recommender Systems: A
Survey of the State-of-the-Art and Possible Extensions”. IEEE Trans. Knowledge and Data
Eng., 17(6): 734–74, 2005.
2. R. B. Allen. “User models: theory, method and Practice”. International Journal of Man–
Machine Studies, 43(11): 27-52, 1990.
3. D. Kalles, A. Papagelis and C. Zaliagis. “Algorithmic aspects of web intelligent systems”. Web
Intelligence, Springer, Berlin, 323–345, 2003.
4. J. L. Herlocker, J.A. Konstan, L. G. Terveen and J. T. Riedl. “Evaluating collaborative filtering
recommender systems”. ACM Trans. on Information Systems (TOIS), 22(1): 5-53, 2005.
5. T. Hofmann. “Collaborative filtering via Gaussian probabilistic latent semantic analysis”. In
Proceedings of the 26th Annual Int’l ACM SIGIR Conference on Research and Development
in Information Retrieval, pp. 145-153, 2003.
6. M. Balabanovic and Y. Sholam. “Combining content-based and collaborative
recommendation”. Comm. ACM, 40(3): 23-43, 1997.
7. Chun Zeng and et al. ”Personalized Services for Digital Library”. In Proceeding of 5th
International Conference on Asian Digital Libraries, pp. 252-253, 2002.
8. G. Ulrike and F. R. Daniel. “Persuasion in recommender systems.” Int’l Journal of Electronic
Commerce, 11(2): 81-100, 2006.
9. S. K. Shinde and U. V. Kulkarni. “A New Approach for on Line Recommender System in
Web Usage Mining”. In Proceeding Inte’l Conference on Advanced Computer Theory and
Engineering, pp. 312-317, 2008.
10. S. K. Shinde and U.V. Kulkarni. “The hybrid web personalized recommendation based on
web usage mining”. International Journal of Data Mining, Modeling and Management, 2(4):
315-333, 2010.
11. K. Yu, A. Schwaighofer and H. P. Kriegel. “Probabilistic Memory-Based Collaborative
Filtering”. IEEE Trans. Knowledge and Data Engineering,16(1) :56-69, 2004.
12. K.W. Cheung and Ch. Tsui, “Extended latent class models for collaborative
recommendation”. IEEE Trans. on Systems, Man and Cybernetics-Part A: Systems and
Humans, 34(1): 143-148, 2004.
13. J. Bezdek. “Pattern Recognition with Fuzzy Objective Function Algorithms”. Plenum Press,
USA, 1981.
14. J.C. Dunn. "A Fuzzy Relative of the ISODATA Process and its Use in Detecting Compact,
Well Separated Clusters". Journal of Cybernetics, 3(3): 32-57, 1974.
15. Y. H. Choa and J. K. Kimb. “Application of web usage mining and product taxonomy to
collaborative recommendations in e-commerce”. Expert Systems with Applications, 26(2):
233-246, 2004.
16. B. M. Kim and Q. Li. “A new approach for combining content-based and collaborative filters”.
Journal of Intelligent Information Systems, 27(1), 79-91, 2006.

17. S. Vucetic. “Collaborative filtering using a regression-based approaches”. Journal of
Knowledge and Information Systems, 7(1): 22-35, 2005.
18. G. L. Somlo and A. Howel, “Adaptive Lightweight Text Filtering”. In proceedings Fourth Int’l
Symp. Intelligent Data Analysis, pp. 53-59, 2001.
19. Z. Huang and W. Chung. “A graph model for e-commerce recommender systems”. Journal of
the American Society for Information Science and Technology, 55(3): 259–274, 2001.
20. D. Billsus and M. “Learning collaborative information filter.” In Proceeding 5
th
International
conference on Machine Learning, pp. 46-54, 1998.
21. C. H. Basu and et al. “Recommendation as classification: Using social and content–based
information in recommendation”. In Proceeding 15
th
International conference on Artificial
Intelligence, pp. 714-720, 1998.
22. Huihong Zhou, Yijun Liu, Weiqing Zhang, and Junyuan Xie. “A Survey of Recommender
System Applied in E-commerce”. Computer Application Research, 1(1): 8-12, 2004.
23. Z. Zhao and S. Bing. “An Adaptive Algorithm for Personal Recommendation”. Journal of
Changchun University, 1(1): 22-29, 2005.
24. L. Teran and Andreas Meier. “A Fuzzy Recommender system for eElections”. EGOVIS 2010,
LNCS, Springer-Berlin, 2(2):62-76, 2010.

Hybrid Personalized Recommender System Using Modified Fuzzy C-Means Clustering Algorithm

More Related Content

What's hot (19)

Viewers also liked (16)

Similar to Hybrid Personalized Recommender System Using Modified Fuzzy C-Means Clustering Algorithm (20)

More from Waqas Tariq (20)

Recently uploaded (20)

Hybrid Personalized Recommender System Using Modified Fuzzy C-Means Clustering Algorithm