MOBILE BUSINESS EXPERT ADVISOR

Danco Davcev, Marjan Arsic and Dalibor Ilievski

Faculty of Electrical Engineering and Information Technologies

University “Ss Cyril and Methodius”, Skopje, Macedonia

Keywords: Mobile Expert, on-demand knowledge, Web services, XML Agents, Fuzzy logic, Pocket PC.

Abstract: In this position paper, we introduce the service which enables consultations by the Experts and/or web

services via mobile devices. The advanced development of wireless networks and mobile devices with

various connection features made a great substrate for the development of services which are based on

immediate response. One such service is a service for on-demand knowledge. A Helpdesk operator can post

a request for consultation to the Experts and/or web services. The goal is to make knowledge available on

demand at any time and any place. This approach gives faster problem solutions, more productive expert

and/or web services, and high availability of the knowledge.

1 INTRODUCTION

The benefit of mobile devices combined with stable

and cheap wireless networks made new kinds of

services to appear. Experts are becoming more

available than ever. Their services can be easy

utilized and customers will have direct answer to the

required question. In some cases (if appropriate) one

of the XML agents communicates with the necessary

Web services to find the solution of the problem.

Exchange of knowledge and consultation process

among customers and available expert authority via

helpdesk operators are very important aspects of

quick problem solution using the mobile devices.

The goal of the mobile technology is to make

information instantly available to customers. The

new business processes can benefit a lot by using

mobile connection between customers and experts.

The exchange of the various content types of media

data (text, drawing, and sound) will enhance the

efficiency of consultations between the customer

and the expert advisor. It gives benefits in reducing

of costs, avoiding future problems or possible

claims, saving time, increasing profit and customer

satisfaction.

There are many desktop applications that

support multimedia communication among

participants. Communication among several

participants that includes multimedia transfer is at

the beginning of the development and utilization in

the sphere of mobile handheld devices. Hence, new

demands are imposed to these devices: greater

processing power to support real time multimedia

transfer (video stream, voice stream, and file

exchange), greater memory space. Modern feature

rich applications demand space on the screen for

displaying all available features, which could be an

issue when handheld devices are used.

A great consideration should be given to the

readability and utilization of small mobile user

interfaces. In this paper, we present our approach in

creating a Mobile Business Expert Advisor (MBEA)

for demanding and exchanging of various content

types and knowledge on mobile devices. The

interface of our MBEA adapts the dimensions and

visibility of the user controls according to the user’s

preferences and utilization of the controls and also,

the current usage of the specific media contents

within the business process. Our knowledge – based

interface is managed by fuzzy logic and many XML

based agents.

The related work in the second section gives

some comments to similar MBEA-s. The

architecture of our Mobile Business Expert Advisor

is elaborated in the third section, which is the main

contribution of this paper. The fourth section

describes the design of MBEA. Implementation and

Evaluation of MBEA are presented in fifth section.

Finally, in the sixth section we conclude the paper.

203

Davcev D., Arsic M. and Ilievski D. (2008).

MOBILE BUSINESS EXPERT ADVISOR.

In Proceedings of the International Conference on e-Business, pages 203-208

DOI: 10.5220/0001913202030208

 SciTePress

2 RELATED WORK

The agent-based approach that uses fuzzy logic to

determine importance of certain information is

elaborated in (

David Camacho, 2001). The agents

described in this paper are based on access to the

Web. We use similar approach to determine

importance of user interface features in regards to

user preferences and media contents used in current

session. However, in our approach is used wireless

network access for communication between

handheld devices.

In (Vlado Glavinic, 2007), an intelligent tutoring

system was presented. They use agents to recognize

the device and the way user device connects to the

global network and according to that information, a

learning content is adjusted to the particular device

and sent to the learner’s device. It differs from our

system because we introduced agents whose purpose

is to intelligently adapt the user interface to the

preferences of each user.

Intelligent Mobile Answering Service is given in

(Business Wire, 2006). Customers can use free-form

questions on any subject with natural language

queries. They can send text messages containing

questions in any form just as easy as they send text

message to their friends and receive a specific

answer to their mobile phones. The natural language

queries allow customers to ask questions in the way

they normally express themselves. Unlike this, our

approach is based on resolving business issues in

various format types by the experts.

In (Lu, L. Kitagata, 2003) an agent based

adaptive user interface control for desktop

applications is presented. Although we share similar

ideas, our approach is multi-agent based adaptive

interface which uses fuzzy logic for handheld

devices.

In (Weichang Du, 2006), collaborative

applications can be built using two types of agent

collaborations, agent communications through

XACL and agent visiting. Agents are represented as

XML entities, not programming language entities.

Secondly, agent hosting services are implemented as

web services with published WSDL, not programs in

certain programming languages with published

APIs. Thirdly, although XML agents’ behaviors

have to be coded in some supported programming

languages, the interactions between agents and hosts

in agents’ behavior code are through invoking local

hosts’ web services, which is neutral to

programming languages and host operating systems.

Although we also use XML entities as agents and

we share similar ideas of using web services (when

appropriate), in our approach we use more flexible

protocol for communication among agents which is

based on fuzzy-based knowledge.

In (Pasquale De Meo, 2007), an XML-based

multiagent recommender system for supporting

online recruitment services is proposed. Although

the main purpose of this system is to provide Online

Recruitment Services, it is also agent and XML

based and as a consequence, it can easily cooperate

with company information systems. The so called

ontology of their user agent stores the profile of a

given user concerning the job search. It is different

from our MBEA because we plan to build a general

purpose advisor system (not only for recruitment

services). For the time being, our objective is to have

a self - adaptive system from the point of view of an

efficient communication between user and the

advisor according to the user's preferences and the

current possibilities of the communication system.

In (Zhiyong Weng, 2007) a feasible framework

that combines agent mobility and intelligence for

consumer-oriented e-business applications is

proposed. This framework complements the current

Web-based systems by adding the wireless channel

of mobile agents. In our work the mobile agents use

an adaptive communication protocol based on fuzzy

logic. In addition, our framework includes web

services.

3 MOBILE BUSINESS EXPERT

ADVISOR (MBEA)

ARCHITECTURE

There are three kinds of users like customers,

helpdesk operator and experts. Helpdesk operator is

the link between experts and customers.

Figure 1: Architecture of the MBEA.

The goal of the helpdesk operator is to response

any kind of requests on demand by the customers in

ICE-B 2008 - International Conference on e-Business

204

appropriate formats (text, draw, VoIP) and to send

the complex requests to the experts and/or web

services which should be resolved by them. Client

applications are implemented on Pocket PCs.

There is only one helpdesk operator and many

experts in one consultation session.

The network communication is realized in a way

that clients send UDP datagrams to the server, and

the server resends the UDP datagrams, according to

the contents of the received message.

The communication protocol between the agents

and the information flow rely on fuzzy logic. In this

case, the fuzzy logic is used for enabling or

disabling (adaptation) some controls depending on

the quality of the service (QoS), i.e. the signal

strength. So, if the signal strength is good, the

features of chat, draw and VoIP will be enabled.

However, for medium signal the VoIP feature will

be disabled. For bad signal, beside the VoIP, the

draw feature should be also disabled, and the only

active communication feature will be chat.

Streams are used for the voice transfer. Desktop

based clients (for both helpdesk operator and

experts) can be also connected to the server.

The Web services are used to find the solution of

the problem using the wireless Internet access. If the

Web services are unable to find the solution, the

Experts are activated. Some types of Web services

are used for the responses. These Web services

support and convert various types of media data.

The experts interface has controls for chat, draw,

VoIP, file upload/download, authority control (VoIP

channel).

The helpdesk operator can give the control over

the interface (or by giving the speaking possibility)

and reclaim the control later. Helpdesk operator has

the role of a moderator in the consultation process.

Figure 1 shows the architecture of the MBEA.

Client applications reside on pocket PCs and they

have three main parts: Helpdesk agent, which

manages operator interaction within the application

and communicates with interoperability agent. VoIP

client serves for the voice transfer, and the client

application represents all the features mentioned

above (chat, draw, voice, file up/download,

communication among clients). Expert Agent

receives and sends the resolved requests using

various file types (like text, image, and voice).

Collaboration Agent calculates the cost of the

corresponding expert’s service. The communication

is realized through wireless area networks

connected to the Internet.

The server contains Interoperability Agent

which redirects the complex request to the Experts

and/or Web services and exchanges reformatted

media types; Passive helpdesk Agent calculates the

helpdesk interface; Media Agent monitors the

media contents (text, image or voice); Experts and

helpdesk preferences are stored in XML data

storage; VoIP Server is responsible for voice

transfer and UDP server application manages the

saving the memory and processing power of the

clients handheld devices, Passive Helpdesk Agents

and Media Agent reside on the server side. The

agents are XML based.

4 MBEA DESIGN

In this paper, we describe our approach in

development of a MBEA by using multi XML

agents and fuzzy logic. Helpdesk agent resides on

the client side, and monitors the helpdesk operator’s

interaction by using the features of the application

and communicates with interoperability agent (e.g.

sends the request to the available competent experts

and/or web services, receives the response in

corresponding format and regulates the customers

billing). The triggers that demand helpdesk agent’s

actions are: drawing, sending chat messages, and

communicating using VoIP. This agent is named

Helpdesk because it helps the customers to simply

resolve their requests. At the start of the

communication session, helpdesk agent

communicates with the Passive helpdesk agent and

interoperability agent that reside on the server. The

Passive agent demands the information from the

Media agent about the type of the files to be used for

the particular communication session. This agent is

named passive helpdesk since it waits to be invoked;

it recalculates the interface structure and sends the

data to the helpdesk agent which finally draws the

interface. The Expert Agent receive the requests in

some format (text, file, voice) from the

interoperability agent, concerns about the solution of

the requests, sends the solutions to interoperability

agent, modifies the expert’s profile, checks the QoS

and billing status.

The main tasks of the interoperability agent are:

receiving the requests from helpdesk agent and

forwarding them to the expert agents and/or web

services. Also it exchanges reformatted data files

with helpdesk agent. Media agent monitors the

folder with the data files for the communication

session, and sends this information to the Passive

helpdesk agent. The Passive helpdesk agent

recalculates the helpdesk interface according to the

MOBILE BUSINESS EXPERT ADVISOR

205

information from the Media Agent, the history of

interface affinities (chat_aff, draw_aff, sound_aff,

QoS) for the particular user and the helpdesk agent’s

message about the use of the features. Then, it sends

the corresponding information (for the type of the

helpdesk interface) to the Helpdesk agent which

adapts the helpdesk interface according to this

information. The collaboration Agent calculates the

time session for realization of the expert task and the

cost of the service

Helpdesk and Expert profiles with personal

information about the qualification and the

knowledge for all of them are stored in the XML

database. Additionally a history of affinities is stored

for the helpdesk operator and the experts.

According to the history, Passive helpdesk agent

adapts the helpdesk interface on the beginning of the

communication session.

There are currently three predefined user

interfaces, which can be preloaded on the client’s

device, according to his preferences and the media

contents for the current communication session. The

dimensions and the positions of the controls would

be stored on the helpdesk’s affinity table on the

server. One kind of an interface displays chat on the

larger part of the screen, the second interface has

larger drawing view on the screen, and the third one

has equal space for the chat and the drawing space.

VoIP feature doesn’t occupy much space on the

screen, so the buttons for voice are always displayed

and they will be active if the Signal Strength is

satisfied, i.e. QoS is strong.

Since there is no universal conclusion for making

decisions of which user interface should be

preloaded, fuzzy logic approach is used to model

that kind of the imprecise information.

The interaction made by the helpdesk operator

using the features of the application (chat, draw and

voice feature), is represented as a vector (chat_aff,

draw_aff, sound_aff). A linguistic variable named

RESULT_PREFERENCE is introduced, and it

accepts values from the set of

terms {increase_chat,

increase_draw, the_same}. This variable represents

the visualization of the helpdesk interface, produced

by (1) the helpdesk operator interaction with the

interface; (2) the contents of media folder and (3) the

quality of service of the signal strength that decides

more accurately which interface to be preloaded.

The class diagram for WiFi signal strength is

shown on Figure 2. Two classes are developed,

WiFiPeek and SignalStrength. The WifiPeek class

implements all the Wi-Fi query related elements.

The class uses the NDIS User mode I/O driver

(NDISUIO) to perform Access Point (AP) scanning.

The GetAdapters function can be used to query

names of network adapters. It calls the built-in NDIS

(not NDISUIO) driver. The function fills a buffer

with adapter names separated by commas. The

RefreshBSSIDs function requests that the driver

initiate an AP survey. It takes one parameter: an

adapter name. The GetBBSIDs function returns the

list of available stations, i.e. peers and Access

Points. The function getPreferredMACAddress

returns the MAC address of the connected

(associated) Access Point.

Figure 2: WiFi signal strength – class diagram.

In order to create the MBEA for calculation of the

user interface visual features, the following fuzzy

variables are defined for this expert system: U_A

(user affinity) which presents the most used feature

by the user, i.e. it has the greatest affinity for the

feature; M_P (media profile) presents the profile of

the interface that should be used according to the

media contents; QoS (Quality of Service) presents

the strength of the wireless signal. Output variable

named I (action for interface adaptation) presents the

necessary interface to be preloaded onto the user’s

device screen, according to the input variables.

The linguistic variable Quality of Service (QoS)

accepts values from the set of terms {Strong,

Medium, Weak}.

The linguistic variable U_A accepts values from

the set of terms {Chat, Draw, VoIP} and it has

normalized values of utilization between 0 and 1

(0% to 100%). It is equivalent to the number of

times of usage of the features chat, draw or VoIP

respectively.

The linguistic variable M_P accepts values from

the set of terms {Text, Drawing, Sound}. It has

normalized values of utilization between 0 and 1

(0% to 100%), which is equivalent to the number of

text, drawing and sound files in the media profile

respectively.

In the table 1 is presented the simulation for

fuzzy linguistic values for U_A, M_P and QoS and

ICE-B 2008 - International Conference on e-Business

206

Figure 3: Agents’ collaboration diagram.

the value of I is calculated, based on the fuzzy rules.

The action needed to adapt the interface is described

with the following XML code like the example 1 of

the Table 1.

</IF>

In the first example from the Table 1, the draw

affinity is more utilized than chat and VoIP and the

media profile has more drawing objects than text or

sound. It is expected to be increased the drawing

interface, but because the signal is weak and the

drawing features are not supported on such signal,

the interface stay unchanged (the_same). The total

numbers of fuzzy rules are 27, the combination of

the values of all linguistic variables. The next

examples are similar like the first one.

Table 1: Some examples for adapting interface using fuzzy

linguistic variables.

The interaction among agents is shown on the Fig. 3.

5 MBEA IMPLEMENTATION

AND EVALUATION

Applications for Wireless MBEA are developed in

C++ Visual Studio .NET 2008 development

framework with using of the MFC (Microsoft

Foundation Classes) library. Operating systems used

on the pocket PCs are Windows CE, Windows

Mobile. The clients which reside on desktop PCs

and the server use the standard .NET Framework.

Operating system for desktop machines on which the

application is practically deployed, is Windows XP

SP2. The interface adaptation is realized on different

screen resolutions.

The two instances of MBEA interface are shown

on Figure 4 for two different screen sizes of Pocket

PC-s.

The initial feedback of MBEA was given by 20

colleagues in ICT. They were asked to use the

system for stock exchange analysis in last week.

General opinion among participants for the interface

Figure 4: Two instances of MBEA.

MOBILE BUSINESS EXPERT ADVISOR

207

usability, functionality and visibility is average

(80%). Interface is functional and suggestions for

interface rearrangment are mostly done according to

the user needs.

The questions that examine interface usability,

functionality, visibility and provide information for

future upgrades are:

1) Are you satisfied with the way of adaptive

presentation?

2) Can you easily select the parameters of MBEA?

3) Were the results of the MBEA solutions are

clearly displayed?

4) How much the MBEA helped in the business

process?

5) Does the MBEA satisfy the quality of the

service?

The questions are answered with “Yes”, “No” or “I

Don’t Know”. The answers are given in the Table 2.

Table 2: Results of the questionnaire.

Question

Q1 Q2 Q3 Q4 Q5

Average

Percent

Yes 17 15 16 14 18 80%

No 2 3 1 4 1 11%

I don’t know 1 2 3 2 1 9%

6 CONCLUSIONS

This paper presents our approach that uses XML

agents and fuzzy logic in order to achieve adaptive

MBEA. We have developed such mobile system

which determines the features of the interface

according to the contents for the specific session.

Fuzzy logic approach is used for the communication

protocol between XML agents and for interface

adaptation. It can be also used to dynamically

reconfigure the interface according to the

preferences and the type of request. In this way, we

proposed an adaptable MBEA for handheld devices

which brings benefits to the developers of

applications for this software environment, to users

of mobile business systems and at the end, to all

mobile device users. General opinion among

participants for the MBEA usability is positive. In

the future work, we plan to provide an detailed

analysis of the user’s feedback.

REFERENCES

David Camacho, Cesar Hernandez, Jose M. Molina, 2001.

“Information Classification Using Fuzzy Knowledge

Based Agents”, Systems, Man, and Cybernetics, IEEE

International Conference; pp. 2575-2580.

Vlado Glavinic, Marko Rosic, Marija Zelic, 2007.

“Agents in m-Learning Systems Based on Intelligent

Tutoring”, Proc. of the HCI Universal Access in

Human-Computer Interaction, Beijing, China, pp.

578-587.

Business Wire, “AskMeNow Launches Intelligent Mobile

Answer Service 'ASKME'”, CNET Networks, 2006.

http://findarticles.com/p/articles/mi_m0EIN/is_2006_

Sept_25/ai_n16838742 [2008]

Lu, L. Kitagata, G. Suganuma, T. Kinoshita, T., 2003.

“Adaptive user interface for multimedia

communication system based on multiagent” Proc. of

the 17th International Conference on Advanced

Information Networking and Applications,

IEEE

Computer Society, pp. 53- 58.

Weichang Du, Hui Li, 2006. “XML Agents Technology

for Building Collaborative Applications”,

Collaborative Technologies and Systems, CTS

International Symposium, Pages: 289 – 297.

Pasquale De Meo, Giovanni Quattrone, Giorgio Terracina,

Domenico Ursino, 2007. “An XML-Based Multiagent

System for Supporting Online Recruitment Services”,

Systems, Man and Cybernetics, Part A, IEEE

Transactions, Volume 37, Issue 4, Pages: 464 – 480.

Zhiyong Weng, Thomas Tran, 2007. “An Intelligent

Agent-Based Framework for Mobile Business”,

Management of Mobile Business, 2007. ICMB 2007.

International Conference, Page: 30

ICE-B 2008 - International Conference on e-Business

208