Intelligent healthcare agent for food recommendation at Tainan City

Intelligent Healthcare Agent for Food Recommendation at Tainan City

Chang-Shing Leea, Mei-Hui Wanga, Wei-Chun Suna, and Young-Chung Changb

aDepartment of Computer Science and Information Engineering, National University of Tainan, Taiwan bAcademia Sinica Grid Computing, Academia Sinica, Taiwan

[email protected]

ABSTRACT—Nowadays, people sometimes eat too much and exercise too little, causing their weight more than their healthy weight range and even developing a disease such as diabetes. In this paper, an intelligent healthcare agent, including an ontology construction mechanism and a food route recommendation mechanism, is proposed for food recommendation at Tainan City. The proposed agent combines the ontology with the fuzzy inference mechanism and the intelligent search mechanism not only to make a guide to Tainan City gourmet but also to display how many calories this gourmet has on the Google Map. In this way, the user can enjoy delicious food while he can stay healthy. The experimental results show that the proposed agent can effectively recommend a personalized schedule of enjoying Tainan City gourmet.

Keywords—Ontology, Healthcare, Agent, Fuzzy Inference, Ant Colony Algorithm

I. INTRODUCTION

Sometimes people eat more calories than they burn, which leads to gaining weight and even suffering from cardio-vascular or chronic diseases. However, it is important for such people to easily count the calories in food that they eat, especially when they take part in outdoor activities. Ontology is with the characteristic of the reusability, which makes it very attractive and powerful for representing domain knowledge [1]. In addition, it is supposed to be shareable across different communities and applications [2]. In the past, some researchers proposed many methods for applying the ontology to different kinds of domain fields. Francisco et al. [3] proposed an ontology-based recruitment system to provide intelligent matching between employer advertisements and the curriculum vitae of the candidates. Liu et al. [4] applied an ontology-based approach to the modeling for the Chinese architecture domain. Hubner et al. [5] proposed an ontology-based search for interactive digital maps. Lee et al. proposed a fuzzy ontology application to news summarization [6], and an ontology-based intelligent decision support agent for Capability Maturity Model Integration (CMMI) project monitoring and control [7]. Ant colony optimization (ACO) takes inspiration from the foraging behavior of some ant species [8]. The algorithm for the ant colony optimization is stochastic search procedures to find good paths through graph [9]. It has caught a numerous researchers’ attention and many successful applications are available. For example, Han and Shi proposed [10] an improved colony algorithm for fuzzy clustering in image segmentation. Zecchin et al. [11] developed parametric guidelines for the application of ACO

algorithm to water distribution system optimization. Ho et al. [12] presented an improved continuous ant colony optimization algorithm to apply to electromagnetic devices designs.

An agent possesses skills, can offer service, is capable of perceiving its environment, or is driven by a set of tendencies [13]. In addition, an intelligent agent differs from an agent in part because of its ability to reason about a task and learn from task performance [14]. Ishida [15] proposed a scenario description language for interactive agents. Lee et al. [16] proposed a genetic fuzzy agent using the ontology model for meeting scheduling system. Besides, they also proposed an ontology-based intelligent healthcare agent to apply to respiratory waveform recognition [17]. Recently, the effects of industrialization on society have resulted in the changes of people’s eating habits, standard of living, and lifestyles so there has been a big increase in the number of people who eat out regularly. Nevertheless, people cannot control the characteristics of food, such as taste, calories per portion, and nutrition when eating out. Under such uncertain conditions, people unconsciously eat high-sugar, high-protein, or even high-fat foods on occasion. Moreover, the problem of population aging around the world has been playing a main factor in emphasizing the importance of the healthcare. In healthcare applications, data, workflow, and logs are often distributed among several heterogeneous and autonomous information system. Kifor et al. [18] proposed a healthcare multi-agent system to analyze the performance and audit provider decisions. In addition, Lange et al. [19] illustrated a framework for integrating heterogeneous ontologies into interdisciplinary, foods-for-health knowledge systems.

The motivation of our work is to provide people, especially suffering from chronic diseases, with an intelligent healthcare agent to make a guide to sampling a gourmet meal and a tip for calories that it has. More specially, the proposed method considers the features of the ontology, the fuzzy inference, the ant colony optimization, and the healthcare agent for food recommendation at Tainan City. Tainan City is an old city located in the southern Taiwan and it has always been famous for its traditional, delicious, and mouthwatering local foods. Undoubtedly, it is a good idea to mix Tainan City gourmet with the core techniques to achieve the goal of healthcare. As a result, we combine Tainan City gourmet with these core technologies to propose an intelligent healthcare for food recommendation to let people both enjoy gourmet and

1465

1-4244-2384-2/08/$20.00 c© 2008 IEEE

eat healthily. The first step in getting a food recommendation is to set up requirements through the Internet. The ontology construction mechanism next maps the requirements into a personal food ontology. The food route recommendation mechanism then implements the fuzzy inference and the intelligent search mechanisms to recommend a personalized food route at Tainan City for the user and this route is stored in the repository.

The remainder of this paper is structured as follows: Section II describes the structure of Tainan City food ontology. Section III presents an intelligent healthcare agent for recommending the food route at Tainan City. The experimental results are shown in Section IV. Finally, some conclusions are drawn in Section V.

II. THE STRUCTURE OF TAINAN CITY FOOD ONTOLOGY

A. Definition of the domain Ontology A novel structure of the domain ontology is proposed in this

paper. Figure 1 shows the structure of the proposed three-layer domain ontology, including a concept layer, a relation layer, and an instance layer [20].

Figure 1. Structure of the domain ontology.

In the concept layer, some concepts, such as “ mCCC ...,,, 21 ”, are defined. Each concept in the concept layer contains a concept name iC , an attribute set

},...,,{ 21 pCCC iiiAAA , an association relation set

},...,,{21 miii CCCCCC RRR , and an instance-of relation set

},...,,{21 inii CICICI RRR . Each attribute has a name and an

attribute value. The relation layer defines some relations. Each relation in the relation layer represents the inter-conceptual relations of the domain ontology. There are two types of relations adopted in the proposed ontology, including the association and instance-of. An association relation represents the inter-relation between two concepts. For example, the association relation between concepts 1C and

4C is 41CCR . The relation between a concept and an instance

is called “instance-of.” The instance layer comprises some instances like “ nIII ...,,, 21 .” Similarly, each instance in theinstance layer is composed of an instance name iI , an

attribute set },...,,{ 21 qIII iiiAAA , and an instance-of relation

set },...,,{21 miii CICICI RRR . The instance-of relation between

instance 1I and concept 1C is 11CIR .

B. Tainan City Food Ontology In this subsection, we describe a scenario for one

application of this paper. Suppose that there is a gourmet who has never been to Tainan City and plans to spend this weekend sampling Tainan City gourmet. He, however, is also concerned about taking in too more calories than his body needs a day after enjoying gourmet. So, he logs on the Tainan City Government’s official tourism Web site (http://tour.tncg.gov.tw/english/index.asp) to learn more about Tainan City, including its local customs and various kinds of foods. After that, he knows that Tainan City is divided into six administrative districts, namely the East District ( ), the

West Central District ( ), the South District ( ), the

North District ( ), the Anping District ( ), and the

Annan District ( ). He also knows that the local traditional food is classified into three types, including main course, snack, and cold drink & ice. To his joy, there are four grades of popularity ranking with gourmet, including most popular, more popular, popular, and less popular, which can help him save a lot of time to search for some of the most popular gourmets. Consider all of the features mentioned above, the domain knowledge of Tainan City gourmet for the proposed agent is able to be represented using the structure of the domain ontology, shown in Figure 1. Figure 2 depicts the architecture of Tainan City food ontology.

Figure 2. Architecture of Tainan City food ontology.

The concepts in the concept layer include “Most Popular,” “West Central District,” “Main Course,” and so on. For example, the association relation between concepts “Main Course,” and “Most Popular” is “Belong to.” The instance-ofrelation between concept “West Central District” and instance “Fuji Meat Dumpings” is called “Locate.” The associationrelation between concepts “Main Course” and “Most Popular” is “Belong to.” The concept “West Central District” has an attribute “Location” and its value is (1300, 850). The instance

1466 2008 IEEE International Conference on Systems, Man and Cybernetics (SMC 2008)

“Fuji Meat Dumplings” has an attribute “Location” whose value is (22.9899, 120.1997).

III. INTELLIGENT HRALTHCARE AGENT

A. Structure of Intellgient Healthcare Agent This subsection utilizes the predefined Tainan City food

ontology to perform the intelligent healthcare agent for food recommendation. Figure 3 shows the structure of the intelligent healthcare agent.

Fuzzy RuleBase

Personal Food Ontology

Fuzzy Inference Mechanism

Intelligent Search Mechanism

Tainan City Food

Information

Location Transfer Mechanism

Tainan City FoodOntology

Concept Selection Mechanism

Relation Generation Mechanism

Instance Mapping Mechanism

Intelligent Healthcare Agent for Food Recommendation at Tainan City

Food Route Recommendation Mechanism Ontology Construction Mechanism Recommended Food Route Repository

User

Domain Experts

FoodRequirements

Figure 3. Structure of the intelligent healthcare agent.

There are two mechanisms, including an ontology construction mechanism and a food route recommendation mechanism in the proposed agent. First, domain experts build the Tainan City food ontology based on the Tainan City food information collected from the Internet. In addition to manual construction, the location transfer mechanism also depends on the Tainan City food information to transfer the value of attribution “Location” of each concept stored in the ontology. Based on the Tainan City food ontology and the user’s requirements, the ontology construction mechanism next carries out the concept selection mechanism, the relation generation mechanism, and the instance mapping mechanism to create the personal food ontology. Then, the food route recommendation mechanism implements the fuzzy inference mechanism to infer the gourmets matched with user’s requirements. Additionally, the intelligent search mechanism performs the ant colony optimization to find a personalized food route according to the top five matched gourmets and then plots this route on the Google Map. Furthermore, the profile of each matched gourmet is simultaneously provided on the constructed Web site. This allows the user to understand whether such a personalized food route can satisfy him or not before starting his gourmet journey to Tainan City. Finally, the recommended food route repository is stored in the repository to allow him to retrieve later.

B. Location Transfer MechanisThe location transfer mechanism is now introduced. The

main functions of this mechanism are to transfer the location attribute’s value of each concept stored in the Tainan City food ontology. According to the geography of the Tainan City’s administrative districts, first we set the attribute-value of “Location” of concepts “East District,” “West Central District,” “South District,” “North District,” “Anping

District,” and “Anna District” to be (1500, 900), (1300,850), (1250,1150), (1400,700), (1000,900), and (900,400), respectively. And then, we analyze the information about the food distribution at Tainan City and further document the statistics in Table I. It indicates that most of the instances stored in the ontology are littered in the “West Central District,” none of which are located in the Annan District. According to the data shown in Table I, the attribute “Location” value for the concepts “Main Course,” “Snack,” “Cold Drink & Ice,” “Most Popular,” “More Popular,” “Popular,” and “Less Popular” is calculated by (1), where x is generated by the value xi and its associated weight wi, y is generated by the value yi and its associated weight wi, Ndenotes the number of administrative districts, xi denotes the attribute value of x axis for the ith concept, yi denotes the attribute value of y axis for the ith concept and wi denotes the associated weight for the ith concept

(1)

1

1

=

=

=

=N

iii

N

iii

wyy

wxx

TABLE I. STATISTICS ON THE DISTRIBUTION OF INSTANCES

Concept Name

Administrative District (%)

East District

West Central District

South District

North District

Anping District

Annan District

Main Course 7 85 0 4 0Snack 0 56 6 13 25 0

Cold Drink & Ice 0 63 0 12 25 0 Most Popular 6 66 0 6 22 0 More Popular 0 96 0 4 0 0

Popular 11 22 11 22 34 0 Less Popular 0 0 0 0 0 0

C. Ontology Construction Mechanism The ontology construction mechanism, including a

concept selection mechanism, a relation generation mechanism, and an instance mapping mechanism, is presented. Based on the Tainan City food ontology and user’s food requirements, first the matched concepts are chosen by the concept selection mechanism. Next, the relation generation mechanism produces the relations of the matched concepts. With the matched concepts and their relations, then the corresponding instances are reflected by the instance mapping mechanism. For example, if there is a scenario that one user wants to taste gourmet foods at some of the “Most Popular” restaurants which serve “Main Course” and are located in the “West Central District.” After analyzing this requirement, the ontology construction mechanism knows that the selected concepts are “Main Course,” “West Central District,” and “Most Popular.” The generated association relations are as follows: (1) “Belong To” between concepts “Main Course” and “Most Popular.” (2) “Locate” between concepts “Main Course” and “West Central District.” Some instances such as “Yangs Superb Dumplings ( )” are mapped. The instance-of relations are below: (1) “Locate” between concepts “West Central District” and “Yangs Superb Dumplings.” (2) “Belong To” between concepts “Most Popular” and “Yangs Superb Dumplings.” (3) “Part of”

2008 IEEE International Conference on Systems, Man and Cybernetics (SMC 2008) 1467

between concepts “Main Course” and “Yangs Superb Dumplings.”

D. Fuzzy Inference Mechanism According to the predefined fuzzy rule base, listed in

Table II, and the personal food ontology, the fuzzy inference mechanism is carried out to infer the context information (CI),the matching degree with user requirements.

TABLE II. FUZZY RULES

Rule Input Fuzzy Variable Output Fuzzy Variable

Semantic Relation(SR) Context Relation(CR) Context Information(CI)

1 VeryLow Low VeryLow 2 VeryLow Medium VeryLow 3 VeryLow High Low 4 Low Low VeryLow 5 Low Medium Low 6 Low High Medium 7 Medium Low Low 8 Medium Medium Medium 9 Medium High High 10 High Low Medium 11 High Medium High 12 High High VeryHigh 13 VeryHigh Low High 14 VeryHigh Medium VeryHigh 15 VeryHigh High VeryHigh

There are two input fuzzy variables, namely semantic relation (SR) and context relation (CR) are adopted. The value of semantic relation is acquired by executing the ant colony optimization, which will be described in the next subsection. First, the fuzzy inference mechanism would retrieve the selected concepts, the generated relations, and the mapped instances from the personal food ontology to calculate the semantic relation and context relation. Figure 4 depicts a simple illustration to describe how to calculate the location distance. The concepts marked with a circle denote the matched concepts with the user’s requirements so “Asong Steamed Bun ( )” is a mapped instance and its location distance is drawn with solid lines in Figure 4. The instance “Zaifa Mean Dumpling ( )” exits instance-of relations between square-marked concepts “More Popular” and “Main Course” so the dotted lines in Figure 4 represent its location distance. Additionally, Figure 4 indicates the longer the location distance, the farer the semantic relation.Next, the context relation is computed. The value of the context relation for the instances “Asong Steamed Bun” and “Zaifa Mean Dumpling” is 3 and 1, respectively. A trapezoidal membership function for fuzzy set FS specified by four parameters FS(x: a, b, c, d) is given in (2) and can be expressed as the parameter set [a, b, c, d].

>≤<−−≤≤<≤−−

<

=

dxdxccdxdcxbbxaabax

ax

dcbaxFS

0)/()(

1)/()(

0

),,,:( (2)

The linguistic terms of input fuzzy variable SR are VeryLow, Low, Medium, High, and VeryHigh. There are three linguistic terms, including Low, Medium, and High, are adopted for fuzzy variable CR. The output fuzzy variable CIhas five fuzzy sets linked: VeryLow, Low, Medium, High, and VeryHigh. The detailed parameters of the membership functions for fuzzy variables are listed in Table III, where mean denotes the value that averages the location distance of all instances stored in the ontology.

More Popular

Location(1304, 844 )

Most Popular

Location(1174, 1269)

Main Course

Location(1306, 849.5)

West Central District

Location(1300,850)

Snake

Location(1235, 861)

Figure 4. Illustration of location distance.

TABLE III. PARAMETERS OF THE MEMBERSHIP FUNCTIONSFuzzy

Variable Linguistic

Term Trapezoidal Membership Function

[a, b, c, d]

SR

SR_VeryLow [mean+50, mean+100, mean+100, mean+150] SR_Low [mean, mean+50, mean+50, mean+100] SR_Medium [mean-50, mean, mean, mean+50]SR_High [mean-100, mean-50, mean-50, mean]SR_VeryHigh [0, 0, mean-100, mean-50]

CRCR_Low [0, 0, 1, 2] CR_Medium [1, 2, 2, 3] CR_High [2, 3, 3, 4]

CI

CI_VeryLow [0, 0, 0, 0.4] CI_Low [0.2, 0.4, 0.4, 0.6] CI_Medium [0.4, 0.6, 0.6, 0.8] CI_High [0.6, 0.8, 0.8, 1] CI_VeryHigh [0.8, 1, 1, 1]

E. Intelligent Search Mechanism

According to the inference results, the intelligent search mechanism chooses the top five matched gourmets as the cities which an ant colony optimization is going to visit in order to find the personalized food route. The state transition rule which means an ant positioned on node r chooses the city s to move is followed the exploitation of a prior and accumulated knowledge about the problem or the exploration of new edges. The applied rule is given by (3) and (4). In addition, the pheromone deposited by ants evaporates with time, so there are two pheromone update rules to allocate a greater amount of pheromone to shorter tours. A local pheromone updating rule is applied while ants construct a solution and is calculated by (5). A global pheromone updating rule is applied only to edges which belong to the best tour and is given by (6). Finally, the personalized food route to taste Tainan City gourmet is displayed on the Google Map. Additionally, the intelligent search mechanism also shows the calories that each matched gourmet has and how many calories a gourmet would get after sampling all of gourmets.

≤⋅= ∈

otherwise,

if},)],([)],({[maxarg 0)(

S

qqururs rJu k

βητ (3)

1468 2008 IEEE International Conference on Systems, Man and Cybernetics (SMC 2008)

where ),(/1),( urur δη = , ),( urδ is the distance between cities r and u , )(rJ k is the set of cities that remained to be visited by ant k positioned on city r, β is a parameter which determines the relative importance of pheromone versus distance and is over zero, q is a random number uniformly distributed in [0, 1], 0q ranging from 0 to 1 is a parameter determining the relative importance of exploitation versus exploration, ),( urτ is the amount of pheromone on edge (r,u), and S is a random variable selected according to the probability distribution given in (6).

otherwise,0

)(if,)],([)],([

)],([)],([

),()(

∈⋅

⋅

=∈

rJsurur

srsr

srpk

rJuk

k

β

β

ητητ

(4)

where ),( srpk denotes the probability with which ant k in city r chooses to move to the city s and τ is the pheromone.

),(),()1(),( srsrsr kτϕτϕτ Δ⋅+⋅−← (5)

where 0),( ττ =Δ srk , ϕ is a pheromone decay parameter between 0 and 1, and 0τ is the initial pheromone concentration.

),(),()1(),( srsrsr kτρτρτ Δ⋅+⋅−← (6)

where

(i)∈

=Δotherwise,0

tour-best-lobal),(if,1),(

gsrLsr gbkτ

(ii) ρ is a pheromone decay parameter between 0 and 1, and

(iii) gbL is the length of the globally best tour from the beginning of the trial.

IV. EXPERIMENTAL RESULTS

We have constructed an experimental Web site at National University of Tainan to test the performance the proposed approach (http://140.133.13.43:8080/TainanFood/index.html). All of the information was collected from the Tainan City Government’s official Web site, including 51 restaurants, of which 27 restaurants serve main course, 16 restaurants serve snack, and 8 restaurants serve cold drink & ice. Figure 5 shows the screenshot of the proposed agent. The first scenario is that one gourmet wants to enjoy some “More Popular” “Main Courses” in the “West Central District” or “South District” or “Anping District.” Then he logs on the constructed Web site to set up his requirements. After waiting for some time, the gourmets matched with his requirements are displayed in Figure 6 (a), whose detailed messages are listed in Table IV. Figure 6(b) shows the recommended food route on the Google Map, which is Yuanhuanding Meat Dumplings ( ) (550 kcal/serving) Yades Chinese

Angelica Duck ( ) (150 kcal/bowl)

Fucheng Shihjingjiou Snack City (

) (305 kcal/bowl) First Class Conquerors Pork Knuckle

( ) (300 kcal/bowl) Rongsheng Snacks (

) (300 kcal/bowl). The second scenario is that another gourmet wants to enjoy some of the “Most Popular” “Snacks” in the “West Central District.” Figure 7(a) shows the matched gourmets and the total calories are 1300kcal if he eats all of the food. The detailed message of each matched gourmet food is also listed in Table IV. Figure 7(b) displays the recommended food route on the Google Map, which directs “Fushenghao ( ) (230 kcal/bowl) Cyuande

Spring Rolls ( ) (300 kcal/serving) Martial Temple

Rice Cake ( ) (230 kcal/bowl) Asong Steamed

Bun ( ) (340 kcal/serving) Fuji Meat Dumplings

( ) (200 kcal/serving).”

TABLE IV. DETAILED INFORMATION ABOUT MATCHED GOURMET FOODNo Instance Name Calories (kcal) Picture

Matched Gourmet Food for No. 1 Scenario

I1Yuanhuanding Meat Dumplings

( ) 550

I2Yades Chinese Angelica Duck

( ) 150

I3Fucheng Shihjingjiou Snack City

( ) 305

I4First Class Conquerors Pork Knuckle

( ) 300

I5Rongsheng Snacks

( ) 300

1605 Matched Gourmet Food for No. 2 Scenario

I6Fushenghao

( ) 230

I7Cyuande Spring Rolls

( ) 300

I8Martial Temple Rice Cake

( ) 230

I9Asong Steamed Bun

( ) 340

I10Fuji Meat Dumplings

( ) 200

1300

Figure 5. Screenshot of the intelligent healthcare agent.

2008 IEEE International Conference on Systems, Man and Cybernetics (SMC 2008) 1469

(a)

.

I4

I3I2

I5

I1

(b) Figure 6. (a) Matched gourmet food and (b) food route for no. 1 scenario.

(a)

I7

I6

I10

I8

I9

(b) Figure 7. (a) Matched gourmet food and (b) food route for no. 2 scenario.

V.CONCLUSION

In this paper, an intelligent healthcare agent is proposed to recommend the route to enjoy Tainan City gourmet. With this agent, the user not only can taste some delicious food but also can know how many calories he takes in. With this provided information, the user can understand if he needs to do more exercise to consume some calories after enjoying gourmet in order to avoid from suffering some diseases caused by obesity. But there are still some problems needed to further study in the future. For example, the genetic learning or neural network will be added to the fuzzy inference to enhance the proposed method. Moreover, the domain ontology for Tainan City food can be applied to other fields.

ACKNOWLEDGEMENTS

The authors would like to thank the financial support sponsored by the National Science Council of Taiwan under the grant NSC95-222-E-024-009-MY2.

REFERENCES

[1] N. Guarino, “Toward a formal evaluation of ontology quality,” IEEE Intelligent Systems, vol. 19, no. 4, pp. 78-79, 2004.

[2] Q. T. Tho, S. C. Hui, and A. C. M. Fong, “Automatic fuzzy ontology generation for semantic web,” IEEE Transactions on Knowledge and Data Engineering, vol. 18, no. 6, pp. 842-856, 2006.

[3] G. S. Francisco, M. B. Rodrigo, C. Leonardo, T. F. B. Jesualdo, and C. N. Dagoberto, “An ontology-based intelligent system for recruitment,” Expert Systems with Applications, vol. 31, no. 2, pp. 248-263, 2006.

[4] Y. Liu, C. Xu, Q. Zhang, and Y. Pan, “The smart architecture: scalable ontology-based modeling of ancient Chinese architectures,” IEEE Intelligent Systems, vol. 23, no. 1, pp. 49-56, 2008.

[5] S. Hubner, R. Spittel, U. Visser, and T. J. Vogele, “Ontology-based search for interactive digital maps,” Intelligent Systems, vol. 19, no. 3, pp. 80-86, 2004.

[6] C. S. Lee, Z. W. Jian, and L. K. Huang, “A fuzzy ontology and its application to news summarization,” IEEE Transactions on Systems, Man, and Cybernetics, part B: Cybernetics, vol. 35, no. 5, pp. 859-880, 2005.

[7] C. S. Lee, M. H. Wang, and J. J. Chen, “Ontology-based intelligent decision support agent for CMMI project monitoring and control,” International Journal of Approximate Reasoning, vol. 48, no. 1, pp. 62-76, 2008.

[8] M. Dorigo, M. Birattari, and T. Stutzle, “Ant colony optimization: artificial ants as a computational intelligence technique,” IEEE Computational Intelligence, vol. 1, no. 4, pp. 28-39, 2006.

[9] M. Dorigo and C. Blum, “Ant colony optimization theory: a survey,” Theoretical Computer Science, vol. 344, no. 2-3, pp. 243-278, 2005.

[10] Y. Han and P. Shi, “An improved ant colony algorithm for fuzzy clustering in image segmentation,” Neurocomputing, vol. 70, no. 4-6, pp. 665-671, 2007.

[11] A. C. Zecchin, A. R. Simpson, H. R. Maier, and J. B. Nixon, “Parametric study for an ant algorithm applied to water distribution system optimization,” IEEE Transactions on Evolutionary Computattion, vol. 9, no. 2, pp. 175-191, 2005.

[12] S. L. Ho, S. Yang, H. C. Wong, K. W. E. Cheng, and G. Ni, “An improved an colony optimization algorithm and its application to electromagnetic devices designs,” IEEE Transactions on Magnetics, vol. 41, no. 5, pp. 1746-1767, 2005.

[13] J. Ferber, “Multi-agent systems,” New York: Addison-Wesley, 1999. [14] H. Chen and J. Yen, “Toward intelligent meeting agents,” IEEE

Computer, vol. 29, no. 8, pp. 62-70, 1996. [15] T. Ishida, “Q: a scenario description language for interactive agents,”

Computer, vol. 35, no. 11, pp. 42-27, 2002. [16] C. S. Lee, C. C. Jiang, and T. C. Hsieh, “A genetic fuzzy agent using

ontology model for meeting scheduling system,” Information Sciences, vol. 176, no. 9, pp. 1131-1155, 2006.

[17] C. S. Lee and M. H. Wang, “Ontology-based intelligent healthcare agent and its application to respiratory waveform recognition,” Expert Systems with Applications, vol. 33, no. 3, pp. 606-619, 2007.

[18] T. Kifor, L. Z. Varga, J. V. Salceda, S. Alvarez, S. Willmott, S. Miles, and L. Moreau, “Provenance in agent-mediated healthcare systems,” IEEE Intelligent Systems, vol. 21, no. 6, pp. 38-46, 2006.

[19] M. C. Lange, D. G. Lemay, and J. B. German, “Perspective a multi-ontology framework to guide agriculture and food towards diet and health,” Journal of the Science of Food and Agriculture, vol. 87, no. 8, pp. 1427-1434, 2007.

[20] N. T. Nguyen, “A method for ontology conflict resolution and integration on relation level,” Cybernetic and Systems, vol. 38, no. 8, pp. 781-797, 2007.

1470 2008 IEEE International Conference on Systems, Man and Cybernetics (SMC 2008)