PARTNER ASSESSMENT USING MADM AND ONTOLOGY FOR

TELECOM OPERATORS

Long Zhang, Xiaoyan Chen

IBM China Research Laboratory, Beijing 100094, China

Keywords: Partner Assessment, Multiple Attribute Decision

Making (MADM), Ontology, Telecommunication.

Abstract: Nowadays, the revenue of telecom operators generated by traditional services declined dramatically while

the value added services involving 3

party value added service providers (partners) are becoming the most

prominent source of revenue growth. To regulate the behaviours of the partners and make the operators be

able to select best service for end users, a flexible partner assessment framework is required. This paper 1)

presents a flexible partner assessment framework based on Multiple Attribute Decision Making (MADM)

method for telecom operators to adapt to the changing requirements of value-added services; 2) proposes

ontology to model the complicated relationship in the assessment factors to achieve high extensibility for

the increasing decision knowledge. From our study, the method adopted and the system proposed can

handle the partner assessment problem and support service selection reasonably in telecom industry.

1 INTRODUCTION

Partner Relationship Management (PRM) is a

business strategy that enables enterprises to manage

and foster profitable partner relationships through

the use of technology. There are many CRM

(customer relationship management) providers that

have incorporated PRM features in their software

applications. PRM can be also considered as a

component of CRM that serves the relationships of

channel partnerships. However, according to Gartner

report, the current PRM research situation is “large

vendors still lag in functionality. Market

consolidation continues, fuelled by vendors that are

combining sell-side commerce with core partner

management functionality.” Thus it is urgent for

enterprise to specify their PRM requirement to

change this situation.

For an enterprise, the most important goal for

PRM is to ac

hieve “win-win”, especially for large

enterprises, such as telecommunication operators.

Now revenue of telecom operators generated by

traditional services, such as local and long-distance

calls, is declining quickly, while data services,

which are also called value-added services, are

becoming the most prominent source of revenue

growth. In data services, telecom operators provide

integrated infrastructures and interfaces, while third-

party service providers which are also called value-

added service providers (VASPs), design and

provide innovative data services to subscribers

through data service delivery platform (DSDP)

which is controlled by telecom operators. Revenue

from data services are shared between VASPs and

operators. NTT DoCoMo has more than 60,000

VASPs. China Mobile, the largest mobile operator in

China, has about 3,000 VASPs and the number is

increasing quickly.

One problem for operators is that these providers

are easy to relapse into malignant competition to

snatch subscribers with shocking means, such as

pricing cheat. Operator’s call centre has to handle

complains from customers. This in a long run will

greatly damages the operator’s image and profit. All

these lead to that the operators have to build some

mechanism to be able to systematically and

scientifically assess the VASPs (partners). Another

important requirement comes from the service

selection requirement. More and more VASPs would

like to wrap their services as web services. The

telecom operator acts as a service agent and

composes the services from its partners (VASPs) to

end users. There may be a lot of candidate services

which are providing same function and suitable for

choosing. The operator has to select among these

candidates according to their historical performance

records and their QoS parameters. Then the operator

232

Zhang L. and Chen X. (2006).

PARTNER ASSESSMENT USING MADM AND ONTOLOGY FOR TELECOM OPERATORS.

In Proceedings of the First International Conference on Software and Data Technologies, pages 232-237

DOI: 10.5220/0001313802320237

 SciTePress

(agent) can determine to invoke the best candidate.

A service assessment mechanism in partner

management system is required.

The situation is that the service assessment

criteria vary a lot. For example, Location Based

Services (LBS) want real time service delivery while

some simple services like weather forecast via short

message (SMS) permit reasonable latency. We may

use delivery time as critical criterion for LBS

services but it is not applicable for weather forecast

services. Meanwhile, telecom operators need to

change the assessment method from time to time

according to the changing market. For example, at

the booming stage of a service, the revenue will be

the operator’s main concern while at a later stage,

customer satisfaction will be as important as

revenue. Therefore, operators need a flexible

assessment framework to adapt to the changing

requirements.

Traditionally, Multiple Attribute Decision

Making (MADM) has been used to select partner in

supply chain. MADM can balance among the

assessment factors and evaluate scores of candidates,

which provides the evidence for decision making.

But it is usually a static selection process. In telecom

industry, the operator needs the ability to

dynamically select proper service among different

candidates in a transaction with an end user. Also,

telecom operators ask for a dynamic assessment

method which will easily connect with operators’

other systems and collect data to perform evaluation

continuously. This paper presents a flexible partner

assessment framework based on MADM method in

which the assessment factors are described using

ontology. The proposed ontology method models the

complicated relationship in the assessment factors to

achieve high extensibility for the continually

increasing decision knowledge for partner

assessment. Domain expert can flexibly design and

revise the structure of assessment ontology, select

and design assessment algorithms, replace the old

ones with new ones in a plug-and-play way. All the

assessment processes are described as a task,

including time-based assessment task and real time

assessment task. Subscriber can subscribe any kind

of assessment task. From our study, the method

adopted and the system proposed can handle the

partner assessment problem and support service

selection reasonably in telecom industry.

The rest of this paper is organized as follows.

Section 2 describes the representation of assessment

model in the assessment framework. Section 3

discussed the ontology for partner relationship

assessment. Section 4 briefly introduces the

assessment framework and a case study. Related

works of MADM, ontology, and the principle of

MADM methodology is described in section 5. We

drew conclusion in section 6.

2 PARTNER ASSESSMENT

MODEL

Assessment is a sub-process of decision making.

Traditionally, the MADA methods which is one

embranchment of Multiple Criteria Decision Making

(MCDM) are used in computer aided evaluation

system to help finish the MADM decision process

when such decision process is complex. The most

frequently used MADA algorithms are AHP

(Analytic Hierarchy Process), TOPSIS (Technique

for Order Preference by Similarity to Ideal

Solution), Weighted Sum Model, Weighted Product

Model, etc. MADM approaches need to define the

assessment factors. In real life, to buy a car, for

example, price, comfortability, security, oil

spending, depreciation, and appearance are all

assessment factors. These assessment factors may

conflict with each other, which make it difficult for

decision makers to balance. MADM provides a

trade-off approach for the decision makers. It

assigns each factor a decision weight and calculates

the weight scores for alternatives or ranks all

alternatives accordingly.

However, this kind of software does not support

continuous decision process which is required by

telecom operators that the assessment and decision

making shall be performed periodically or on ad-hoc

requests. They are also difficult to integrate with

other systems in an enterprise environment and

effectively manage the large volume of intermediate

assessment results. Here we designed an assessment

framework, which can not only support continuous

assessment based on the existing MADA algorithms,

but also provide effective assessment result

management. First, it is necessary to define the

assessment model.

Definition 1: A Candidate c is an entity which is

evaluated. It is same with alternatives in common

MADM literature. A candidate c relates to

relationship model M, and has instances {I

, I

, … ,

} to be assessed by evaluator E in assessment task

T respectively.

The relationship model, instance, and evaluator

will be defined below. Only the candidates who

share the same relationship model and evaluator can

be evaluated together in a specific evaluation task.

PARTNER ASSESSMENT USING MADM AND ONTOLOGY FOR TELECOM OPERATORS

233

The candidate type and its characteristics can be

defined by user.

Definition 2: A Relationship Model M is a tree,

whose nodes are assessment Factors {mf

, mf

, ... ,

, lf

, ... , lf

}, where mf is middle factor node

in the tree while lf is leaf factor node. Relationship

model is the basis of the assessment framework.

Definition 3: Priority P={p

, p

, … , p

} is a

domain expert knowledge on the factors. Different

evaluators use different forms of priority. For

example, the AHP evaluator use pair-wise

comparisons to get each leaf factor’s weight w

, and

use W={w

, w

, … , w

} as its priority.

Definition 4: An Evaluator e is a model on how to

evaluate instances using a relationship model. Each

evaluator involves a relationship model, a priority,

and an evaluation algorithm. For example, the AHP

evaluator will use the priority multiplies the

instance: {w

, w

, … , w

} as evaluator model.

Definition 5: Assessment Result S is the score of

an instance being evaluated with an evaluator, which

can be gotten by for AHP evaluator.

∑

lfwS

An assessment is a one-off operation performed

on the instances. Its input is the instance data

,…,I

} and its required evaluator E. E can be

AHP evaluator, TOPSIS evaluator or others. The

output is the assessment Result S, and the rank of

each candidate c according to assessment Result S.

An assessment task T is an Assessment set.

Assessment tasks can be classified as time-based

evaluation and event-based (real time) evaluation.

The task engine gets each task from knowledge base,

parse and execute it. This will be described in

section

For time-based evaluation task, the parameters

include assessment objects (candidates), time

duration, evaluation frequency, evaluator and the

mode to get assessment result. The instance data for

a specific candidate c will be obtained every fixed

sample time interval, and the assessment will be

done periodically. For an event based evaluation

task, the parameters include evaluation objects

(candidates), time duration, the factors whose update

will trigger assessment engine to do assessment,

evaluator and the mode to get assessment result.

3 ONTOLOGY FOR PARTNER

ASSESSMENT

Making explicit domain assumptions underlying an

implementation makes it possible to change these

assumptions easily if our knowledge about the

domain changes. Hard-coding assumptions about the

world in programming-language code make these

assumptions not only hard to find and understand

but also hard to change, especially for someone

without programming expertise. In addition, explicit

specifications of domain knowledge are useful for

new users who must learn what terms in the domain

mean.

We need a flexible middle layer. This will make

it possible that the system structure and algorithms

running upon can be plug-and-play. Separating the

domain knowledge from the operational knowledge

is another common use of ontologies. For example,

we can describe a task of configuring a product from

its components according to a required specification

and implement a program that does this

configuration independent of the products and

components themselves.

Ontology has been playing an increasingly

important role in many applications, because it

provides: (1) a shared and common understanding of

the knowledge domain that can be communicated

among agents and application systems, and (2) an

explicit conceptualization that describes the

semantics of the data (Fensel et al, 2000). These two

properties of ontology are crucial in modelling the

complicated relationships among assessment factors

and achieving a high extensibility for the continually

increasing decision knowledge for partner

relationship assessment. We developed a business

partner evaluation ontology for the telecom domain

to support the decision making process.

3.1 Partner Assessment Ontology

Definition

Part of the upper level of the evaluation ontology

developed for PRM is showed in the Figure 1, where

the core concept Evaluation Profile is the common

superclass to describe all kinds of business partner

evaluation.

Figure 1: Upper level part of the ontology.

ICSOFT 2006 - INTERNATIONAL CONFERENCE ON SOFTWARE AND DATA TECHNOLOGIES

234

z Business Partner: the object to be evaluated. It

has name and other attributes.

Figure 3: “Service” in Provider Evaluation ontology.

z Evaluation Dimension: a kind of measure for

evaluating a business partner. The evaluation

dimension can be divided into Atomic

Dimension and Complex Dimension. The

Complex Dimensions are composed of (atomic

dimension or complex dimension) dimensions.

z Metric: a standard of measurement. Metric is a

common superclass for all other metrics and

has related property metricUnit, metricValue

and metricName. Metrics can be divided into

Atomic Metrics and Complex Metrics. The

atomic metrics are directly measured by

corresponding observers. The complex metrics

are composed of other (atomic metric or

complex metric) metrics.

z Evaluation Category: describes categories of an

evaluation on the bases of some classification.

By applying sub-classing to the concepts in the

upper ontology, we then developed the Provider

Evaluation sub-ontology, partially showed in Figure

2. We investigated partner assessment methods of

some operators. Based on our analysis, we propose

the following dimensions for partner assessment.

z Service. “Service” is a complex dimension and

can be divided into “Service QoS”, “Service

Interface” and “Service Function” dimensions.

The metrics for “Service QoS” are “Ratio of

download overtime times”, “Ratio of response

Overtime Times”, “Definitions and

Readability”, and “Convenience for use”. The

metrics for “Service Interface” are “Service’s

physical connection”, “protocol of interface”

and “possibility of exceptions”. Figure 3

depicts the definition of “Service”.

z Revenue. It includes “Revenue Ration”,

“Revenue Incremental Ration”, “User Number

Ratio”, “User Incremental Ratio”, “Homepage

Visit Ratio”, and “Homepage Visit Increment”.

z Management. It includes “Follow-up Required

Propagandize”, “Result in negative Report

from Media”, “Violate Cryptic Agreement”,

“Provide Charge Agent Service”, “Customer or

Price Cheating”, “Push Advertisement without

Customer Permission”, and “Cooperation

Satisfaction”.

z Customer QoS. It includes “Complaining

Ratio”, “Complaining Processing Ratio”,

“Complaining Processing Efficiency”, and

“Customer Call Switch-on Ratio”.

3.2 MADM based on Partner

Assessment Ontology

Many MADM algorithms compose the evaluation

factors as a tree. Usually, the evaluation factor tree

comes from information system. With the

development of knowledge base technology, an

enterprise information schema will be or has been

described by ontology as a reusable asset. Ontology

usually has a graph structure, while evaluation

factors employ a tree. In our previous work

(Nanavati et al, 2005), a method was proposed to get

MADM required tree structure from ontology with

graph structure, which will be an important

processing for building a MADM based evaluation

system. The transformed ontology tree will then act

as the Relationship Model defined in section 2. For

example, in Figure 3, assume the sub-ontology is a

tree structured one after transformation. The leaf

factors are assigned with weights w

, w

…. The weights of middle factors can be calculated

recursively. For example, w

;

Currently, ontology storage and query is being

widely investigated in the semantic web community

(Beckett, 2003; Alexaki et al, 2001). Moreover,

many researchers are working on evaluating the

performance of ontology repositories (Guo et al,

2005; Tempich and Volz, 2003). The candidates and

instances defined in section 2 can be stored and

queried using the existing techniques.

Figure 2: Provider Evaluation sub-ontology.

PARTNER ASSESSMENT USING MADM AND ONTOLOGY FOR TELECOM OPERATORS

235

4 PARTNER RELATIONSHIP

ASSESSMENT FRAMEWORK

AND CASE STUDY

Based on the above definition of assessment and

ontology, we design the architecture of assessment

framework as shown in Figure 4. There are three

roles of external actors of the assessment

framework. The actors may be from telecom

operation departments: 1) Domain Expert who

defines relationship model M, and the specific

evaluator E. 2) Data Operator who inputs

information of candidate according to requirement,

and inputs information of instances of the candidate

as defined in section 2. 3) Subscriber who subscribes

assessment task T, gets assessment result S using the

mode predefined, and starts the assessment engine to

do assessment.

Usually, the instance data are gotten from outer

systems, such as Call Center, Billing System. The

Instance Data Receiver supports both pull and push

modes to get data and store into knowledge base by

Semantic Framework.

z Relationship Model Designer

A domain expert builds its own relationship

model by using existing models or creates a brand

new one. In MADM approach, the relationship

model can be depicted as a tree. Ontology is used to

describe the tree and is stored into knowledge base.

z Algorithm Framework

All the evaluation MADM algorithms can be

plugged into the Algorithm Framework. When a

domain expert decides to use a specific MADM

algorithm such as AHP algorithm, the expert can

tailor the relationship model to decide the

assessment factors, and select or designate each

factor’s weight.

z Task Manager and Task Engine

All the evaluation tasks are designed and

managed in Task Manager: get evaluation tasks

from the database and manage the execution. Task

Engine is the place where a task is actually

evaluated.

z Semantic Framework

With the ontology engaged in the framework, we

achieve a flexible framework and can flexibly define

and change the evaluation algorithms. The Semantic

Framework is responsible for storing data into the

knowledge base and evaluating requested queries.

Based on the description, we implemented a

partner relationship management prototype system.

Preliminary experiments show that the method

adopted and the system can handle the partner

relationship assessment and give guidance of service

selection reasonably.

Here a case study is used to show the usage of

PRM system. A mobile operator can offer a “Ticket

Master” like service. The process is depicted in

Figure 5.User John sends request to his mobile

operator through short message, Web portal, or IVR

(Interactive Voice Representative) to book movie

tickets for a theatre in his vicinity. The tickets are

booked, and the theatre is informed of the booking.

Since John has an account with the operator, his

address is looked up. Based on the locality, the

corresponding courier is selected and John’s address

is passed on to the courier. The courier gets driving

directions from a mapping service which delivers the

ticket and (optionally) collects the payment. The

payment may be collected on delivery or billed to

John in his next billing cycle.

Note that “Deliver Tickets” and “Collect

Payment” are not web services, but are physical

activities that are reflected in the electronic world by

their confirmation. “Book Tickets” happens at the

mobile operator. “Inform Theatre” happens at the

Theatre. “Address Lookup” happens at the Directory

Service Provider (DSP). “Select Courier” happens at

the mobile operator. “Driving Directions” happens at

the Mapping Service Provider (MSP). “Payment

Confirmation” happens at the mobile operator.

Figure 5: A case study.

Figure 4: Architecture of the Evaluation Farm.

ICSOFT 2006 - INTERNATIONAL CONFERENCE ON SOFTWARE AND DATA TECHNOLOGIES

236

A single mobile operator may have 5 theatres, 3

couriers, 3 DSPs, 2 MSPs and 3 banks registered as

service providers. Therefore, partner evaluation and

service selection is possible. The relationship

ontologies for theatres, couriers, DSPs, MSPs and

banks should be defined according to the methods

described in section 3.

5 RELATED WORKS

MADM refers to the problem of selecting among

alternatives associated with multiple, usually

conflicting, attributes where the decision maker’s

preference information is often used to rank

alternatives. As a branch of decision making

method, MADM has gained wide usage in

management and engineering. For example, Aura

Reggiani uses AHP and TOPSIS to evaluate a set of

a priori selected airports alternatives for airline

(Janic and Reggiani, 2002). Maggie C.Y. Tam and

V.M. Rao Tummala use AHP to select the vendor

for a telecom system (Tam and Rao, 2001). They

investigated the feasibility of applying the AHP in

vendor selection for a telecom operator to improve

the group decision making by a more systematic and

logical approach. Work in (Ceccaroni et al, 2004)

proposes the OntoWEDSS system which uses

ontology to improve the diagnosis of faulty states of

a treatment plant. The system supports wastewater-

related complex problem-solving, and it facilitates

knowledge modelling. Work in (Li et al, 2001) uses

ontology to describe the competencies of an

enterprise based on which a decision support system

for enterprise bidding is built.

This paper distinguishes itself that it investigated

a specific industry – telecom and provides a solution

based on MADM for operators to assess their

partners (VASPs) effectively. The ontology to model

the complicated relationship in the assessment

factors helps achieve a high extensibility for the

increasing decision knowledge for partner

assessment. The proposed system is easy to integrate

with other telecom systems.

6 CONCLUSION

The paper presented a flexible partner assessment

framework based on Multiple Attribute Decision

Making (MADM) method for telecom operators to

adapt to the changing requirements of value-added

services, and proposed to use ontology to model the

complicated relationship in the assessment factors to

achieve high extensibility for the continually

increasing decision knowledge for partner

assessment. Preliminary usage of our prototype

system showed that our approach was practical to be

used in telecom industry. The approach would

dramatically improve customer experience and the

quality of services. Furthermore, this will

significantly improve the competing ability of

telecom operators and increase their marginal profit

especially with the 3G and NGN (Next Generation

Networks) bloom the data services.

REFERENCES

Tam, C.Y. and Rao, V.M., 2001. An Application of the

AHP in Vendor Selection of a Telecommunications

System. In the Int’l J. of Management Science, Omega

29, pp. 171-182.

Janic, M. and Reggiani, A., 2002. An Application of the

Multiple Criteria Decision Making Analysis to the

Selection of a New Hub Airport. In European J. of

Transport and Infrastructure Research, no. 2, pp. 113.

Fensel, D., Horrocks, I., Harmelen, F., Decker, S.,

Erdmann, M. and Klein, M., 2000. OIL in a nutshell.

In EKAW 2000, pp. 1-16.

Beckett, D., 2003. Scalability and Storage: Survey of Free

Software/Open Source RDF storage systems, available

at www.w3.org

Alexaki, S., et al, 2001. On Storing Voluminous RDF

Description: The case of Web Portal Catalogs, In Proc.

of the 4th Int’l Workshop on the Web and Databases

(WebDB).

Guo, Y., Pan, Z. and Heflin, J., 2005. LUBM: A

Benchmark for OWL Knowledge Base Systems. In J.

of Web Semantics, Vol 3, Issue 2.

Tempich, C. and Volz, R., 2003. Towards a benchmark for

Semantic Web reasoners – an analysis of the DAML

ontology library. Workshop on Evaluation on

Ontology-based Tools, ISWC2003.

Ceccaroni, L., Cortes, U. and Sanchez-Marre, M., 2004.

OntoWEDSS: an ontology-underpinned decision-

support system for wastewater management.

Environmental Modelling and Software 19(9): 785-

797.

Nanavati, A., Yu, X., Chen, X. and Chen, Y., 2005. A

Method for Ontology Transformation for Multiple

Attribute Decision Making. Technical Report, IBM

China Research Lab, June 2005.

Li, Y., Huang, B., Liu,W., Gou, H. and Wu, C., 2001.

Ontology for modeling and analyzing of enterprise

competence. IEEE Int’l Conference on Systems, Man,

and Cybernetics.

PARTNER ASSESSMENT USING MADM AND ONTOLOGY FOR TELECOM OPERATORS

237