GENERIC ARCHITECTURE FOR INCOORPORATING

CLUSTERING INTO e-COMMERCE APPLICATIONS

Anastasios Savvopoulos and Maria Virvou

Department of Informatics, University of Piraeus, Karaoli & Dimitriou St, Piraeus, Greece

Keywords: User modelling, e-Commerce, e-Shopping, Adaptivity, Clustering, Stereotypes, Animated agents.

Abstract: Today product recommending applications use many techniques in order to achieve personalization. These

techniques may prove successful but lack in portability. This means that it is very difficult to apply the same

architecture and techniques that have been used on one system to a totally different one. In this paper we

propose a generic architecture that can be used to achieve personalization to a product recommending

system. The main advantage of this architecture is that every system can use it, even if it is built on php,

asp.net or a different web technology. In this paper we present a case study that we applied this architecture.

This case study proves the independency of our architecture and that it can be applied easily to any kind of

remote recommending system

1 INTRODUCTION

E-commerce applications have become very popular

since they provide easy access to all kinds of

products. However, most of existing applications are

generic and do not address specific needs,

preferences and attributes of individual customers. A

remedy to this problem can be achieved by web

personalization techniques. As De Pessemier et.al.

(Pessemier et.al. 2008) suggests, new technologies

such as Internet, iDTV, and mobile applications

create the possibility to advertise in a different, more

attractive manner than the traditional commercial

breaks. This leads to a rise of interactive

commercials on websites, banners on the internet,

and commercials on mobile devices. Thus the

market moves towards more converged architectures

corresponding to different types of mediums such as

mobile phones, internet. Despite the fact that there

are many techniques in order to achieve

personalization, there is a lack in the effort to

produce frameworks that can be applied to any

product recommending application without

concerning the product that sells or the medium that

this application uses. Also as Veruska R. Aragao

suggests, (Aragao et.al. 2001) there is no widely-

available mechanism to allow users to personalize

their interaction with web data and services

meaning that as years pass the need for general

personalization architectures is becoming more and

more imperative. The difficulty of making such a

framework is high and it’s reinforced by the fact that

product brokering requires assisting users in finding

information in a complex multidimensional space

(Pu & Faltings, 2002). In this paper we present a

generic architecture for e-commerce applications

that incorporates a clustering algorithm in order to

create groups of similar users, concerning their

needs and interests.

2 RELATED WORK

There are many e-shops applications that try to make

recommendations using many techniques. These

techniques usually involve the construction of user

models that are either based on explicit user

information or on data about the user behaviour that

is collected implicitly by the system. An interesting

approach in the field of e-commerce has been made

by Choi, et al. (Choi et.al. 2006). They chose a

multi-attribute decision making method to find

similar products. The procedure of finding a similar

product is based on a weighted attribute theory that

can fill incomplete specifications of a similar

product if these specifications do not exist.

Another interesting approach in the same field

that uses clustering techniques in order to group

products is the system created by Guan et al. (Guan

et.al., 2005) . In their system they use an explicit

427

Savvopoulos A. and Virvou M.

GENERIC ARCHITECTURE FOR INCOORPORATING CLUSTERING INTO E-COMMERCE APPLICATIONS.

DOI: 10.5220/0001840304270430

In Proceedings of the Fifth International Conference on Web Information Systems and Technologies (WEBIST 2009), page

ISBN: 978-989-8111-81-4

method of ranking to acquire generic attributes from

products and then cluster new attributes into the

different groups of generic attributes using the k-NN

algorithm. A very interesting technique also, based

on a rating system has been conducted by Q. Li and

B. M. Kim (Li & Kim, 2004). The system acquires

rates and then calculates fuzzy inferences and

extracts similarities between users. Their method

proved very successful according to the evaluation

presented. Another interesting research has been

proposed by Kazienko and Kolodziejski (Kazienko

& Kolodziejski, 2005). Their system is called

WindOwls and is a recommending system that uses

user modelling techniques to propose products to

individual users. Windowls uses association rules to

calculate weights in order to group acquired tastes

together.

Despite the fact that all the above systems

provide users with recommendations, they are so

domain and problem depended that lose the ability

to be applied with easy in different fields or

products. In this way every time a new application is

built a new architecture must be constructed from

scratch in order to address the specific application’s

problems. The framework presented here is product

and domain independent. The main advantage is that

it can be applied on any product recommending

application without consideration of the domain or

the products used such as personal computers,

mobile phones or even cars.

3 THE PROPOSED

ARCHITECTURE

The main architecture of our framework is divided

into two general sections (figure 1). The first section

contains elements that do not interact with the users

directly and the second section contains elements

that users can understand and interact with. The

elements included in the first section are: Explicit

Information User Profiles, Observing Behavior

Agent, Clustering Algorithm Process, Double

Stereotypes and User Model Server. The Explicit

Information User Profiles element contains all the

information in a database that users have provided

the system in an explicit way. Either, by answering

interest questions or rating products. The next

element is Observing Behaviour Agent. This

element plays a key role in the construction of the

User Model and contains all the information about

the user interactions with the system. Also observes

users’ actions throughout the usage of the system.

Figure 1: The General Architecture.

For example, this element contains information

about the categories a user has visited, products that

s/he visited, products that s/he moved in or out of his

cart and products that s/he bought. The Clustering

Process Algorithm element contains the clustering

algorithm that the system uses to group similar users

and extract representative users of these groups. The

clustering algorithm takes as input the statistical data

of all the explicit and implicit information that the

system measures about users. Every category and

product characteristic is a feature measured by the

algorithm. The clustering algorithm processes these

data and provides the system with groups based on

similarity. From these groups representative feature

vectors are extracted. After this procedure the

Double Stereotypes element calculates dynamic

double stereotypes from these representatives. These

stereotypes follow a general to specific hierarchy,

meaning that the system constructs a low number of

generic stereotypes at first and then continues to

construct more specific stereotypes until it reaches a

certain point of complexity. These four elements

mentioned above contribute to the construction of

the User Model Server component. The User Model

Server component contains all the information,

implicit and explicit, about a specific user, the

stereotype that s/he belongs to and his/her more

similar representatives and manages all the above

elements. The second section contains elements

interacting directly with the user. The Incremental

Initialization Process element acquires information

from the user model and tries to provide the best

recommendations to new users or users that the

system has little information about. This element

uses mostly stereotypic information from the Double

WEBIST 2009 - 5th International Conference on Web Information Systems and Technologies

428

Stereotypes element and makes assumption about

where the user should belong to according to the

moves that s/he made so far. The Recommendation

element communicates with the User Model element

and provides the users with recommendation about

products and system usage. The Recommendation

element contains all system recommendations about

products, mistakes or other recommendations in a

database. This element can use many techniques in

order to make recommendations such adaptive

hypermedia, dynamic annotations and the animated

agent. The Recommendation element also takes

feedback from the users and provides the User

Model with more useful information. The next

element is the Animated Agent. It’s a system

component that manages an animated agent that can

help users throughout the navigation of the system.

This agent can provide useful information about the

usage of the system and make recommendation

about products by acquiring information from the

user model. Next, we have the User Interface

element. We use a dynamic user interface that not

only adjusts to the medium used automatically, but

also changes according to the users’ interests. The

Recommendation and Incremental Initialization

elements can change the User Interface according to

the User Model of every user. In this way if a user

uses a mobile phone or an interactive tv the

experience would be different. We must note here

that despite the fact that user interface experience is

different for every medium, it contains its basic

characteristics in order to avoid user confusion. Last

but not least, we have the Adaptive Hypermedia

element that is used to annotate user interface

elements according to the users’ needs or

preferences. All these elements communicate with

each other using the User Model Server as passage

from one to another. The diagram below shows a

schematic representation of the proposed

architecture and how the above elements are

connected together. The main interaction between

the user and the intelligent system is made through

the dynamic user interface.

4 CASE STUDY

Our first case was video store application called

Vision.Com (Virvou et.al. 2007). Vision.Com is an

e-commerce video store that learns from customers’

preferences. Its aim is to provide help to customers

choosing the best movie for them.

For every user the system creates a different

record at the database. There are two types of

information saved for every user, the explicit and

implicit. The explicit information is saved on the

Explicit Information User Profiles and the implicit

are saved by the Observing Behaviour Agent. In

Vision.Com every customer can visit a large number

of movies by navigating through four movie

categories.

All navigational moves of a customer are

recorded by the system in the statistics database by

the Observing Behaviour Agent. In this way

Vision.Com saves statistics considering the visits in

the different categories of movies and movies

individually, movies that were moved to buyers’ cart

and bought movies. Each user’s action contributes to

the individual user model by implying degrees of

interest into one or another movie category or

individual movie. Apart from movie categories that

are already presented, other movie features that are

taken into consideration by Vision.Com are the

following: price range, leading actor and director.

The price of every movie belongs to one of the five

price ranges: 20 to 25 €, 26 to 30 €, 31 to 35 €, 36 to

40 € and over 41 €.

Vision.Com incorporates an AIN clustering

algorithm (Cayzer & Aickelin, 2002), (Morrison &

Aickelin, 2002). This algorithm was built by D.N

Sotiropoulos et.al. (Sotiropoulos et.al 2006) This

algorithm takes as input feature vectors that contain

all the statistical data of all users. Then AIN

algorithm processes this information and provides

the system with representatives. This process is

being conducted by the Clustering Process Element

of our architecture. We used these representatives in

order create double stereotypes that are saved in the

Double Stereotypes element of our general

architecture. The stereotypes concern both users and

movies. In order to find similar movies to a users’

interest through the stereotypical information we use

the Euclidean distance. The hierarchy of the

stereotypes follows the complexity of the

stereotypes. Levels 1 are the more general

stereotypes and level 5 the more complex ones,

depending on the information the system has about a

user. After level 5 the system uses an individual user

model about this user. This process also helps the

Incremental Initialization Process to provide better

results about new users. Moreover, Vision.Com

uses an Animated Agent (figure 2) to inform and

help the users and provide general

recommendations.

GENERIC ARCHITECTURE FOR INCOORPORATING CLUSTERING INTO E-COMMERCE APPLICATIONS

429

Figure 2: The animated agent.

5 CONCLUSIONS

In this paper we proposed a generic architecture that

can be used to achieve personalization to a remote

recommending system. The main advantage of this

architecture is that its medium independent meaning

that every system can use it, even if it is built on a

mobile phone, a pc or a tv. We presented two case

studies that we applied this architecture that use

entirely different products, mediums and kinds of

recommendations. These two cases prove the

independency of our architecture and how easily it

can be applied to any kind of remote recommending

system.

ACKNOWLEDGEMENTS

Travel fund support for this work was provided by

the University of Piraeus Research Center.

REFERENCES

Cayzer S., Aickelin U, 2002. A Recommender System

based on the Immune Network. In Proceedings of the

Congress on Evolutionary Computation

Morrison T., Aickelin U, 2002. An Artifcial Immune

System as a Recommender for Web Sites. In

Proceedings of the 1st Conference on ARtifcial

Immune Systems (ICARIS-2002). Canter-bury UK.

161-169

Choi S. H., Kang S., Jeon Y. J, 2006. Personalized

recommendation system based on product

specification values. Journal of Expert Systems with

Applications vol. 31, Elsevier. 607–616

Guan S-U., Tai Kheng Chan T. K., Zhu F, 2005.

Evolutionary intelligent agents for e-commerce:

Generic preference detection with feature analysis.

Journal of Electronic Commerce Research and

Applications vol. 4, Elsevier 377–394

Kazienko P., Kolodziejski P, 2005. WindOwls – Adaptive

System for the Integration of Recommendation

Methods in E-Commerce. AWIC 2005, Lecture Notes

in Artificial Intelligence vol. 3528, Springer-Verlag

Berlin Heidelberg 218–224

Maybury M., Greiff W., Boykin S., Ponte J, Mchenry C.,

Ferro L, 2004. Personalcasting: Tailored Broadcast

News. Journal of User Modeling and User-Adapted

Interaction vol. 14, Kluwer Academic Publishers,

Netherlands 119-144.

Goren-Bar D, Glinansky O, 2004. FIT-recommending TV

programs to family members. Journal of Computer

and Graphics vol. 28, 149-156.

O’ Sullivan D., Smyth B., Wilson D. C., Mcdonald K.,

Smeaton A, 2004. Improving the Quality of the

Personalized Electronic Program Guide. Journal of

User Modeling and User-Adapted Interaction vol. 14,

Kluwer Academic Publishers, 5-36.

Virvou M, Savvopoulos A, 2007. An Intelligent E-

shopping Application that provides Recommendation

In 19th IEEE International Conference on Tools with

Artificial Intelligence ICTAI 2007, 29-31 October,

Patras, Greece.

Pessemier T.D., Deryckere T., Vanhecke K., Martens L.,

2008. Proposed Architecture and Algorithm for

Personalized Advertising on iDTV and Mobile

Devices, In Journal of IEEE Transactions on

Consumer Electronics.

Pu P., Faltings B., 2002. Personalized Navigation of

Heterogeneous Product Spaces using SmartClient, in

Proceedings of the International Conference on

Intelligent User Interfaces.

Aragao V.R., Fernandes A.A., Carole A. 2001. Goble,

Towards An Architecture For Personalization And

Adaptivity In The Semanticweb. In Int. Conf. on

Information Integration and Web-based Applications

& Services, Linz, Austria, Sept.

Li Q., Kim B.M., 2004. Constructing User Profiles For

Collaborative Recommender System, Lecture Notes in

Computer Science, Springer-Verlag, pp 100-110.

Sotiropoulos D.N., Tsihrintzis G.A., Savvopoulos A.,

Virvou M., 2006. A Comparision of Customer Data

Clustering Techniques in an e-Shopping Application,

Adaptive Hypermedia and Adaptive Web-Based

Systems, June 20 - 23, 2006

WEBIST 2009 - 5th International Conference on Web Information Systems and Technologies

430