DEVELOPING OPTIMIZATION ALGORITHM USING ARTIFICIAL BEE COLONY SYSTEM

Saif Mahmood Saab Arab Academy for Banking and Financial Sciences, AmmanJordan saif.alsaab@gmail.com Dr. Nidhal Kamel Taha El-Omari The World Islamic Sciences & Education University (WISE) ), AmmanJordan omari_nidhal@yahoo.com Dr. Hussein H. Owaied Middle East University for Graduate Studies, AmmanJordan howaied@meu.edu.jo

ABSTRACT More recently, computer scientists have found in the study of social insects behavior a source of inspiration for the design and implementation of optimization algorithms. Particularly, the study of bee colonies behavior turned out to be very attractive to develop problem solving algorithms. However, this paper introduces a new algorithm that are inspired by bee colonies. It presents two unsupervised scenarios for two kinds of bees which are scouts and forgers in order to identify their activities that used by the bees during the food forging, These activities have been analysis and designed as proposed new optimization algorithms as the functions to be used by the processing engine of the artificial bees colony system. The proposed algorithm can be applied to dynamic problems in real variables, stochastic problems, multi-targets and parallel implementations, and any areas which are not required supervision. Keywords: Artificial Intelligence (AI), Expert systems, natural system, Problem Solving algorithms, Swarm Intelligence (SI).

INTRODUCTION

Since ever, many natural systems of the most creatures in the world are very rich topics for the scientific researchers. However, a simple individual behavior can cooperate to create a system able to solve a real complex problem and perform very sophisticated tasks. In realty there are many patterns of such systems like ant colonies, bird flocking, fish shoaling, animal herding, bacterial growth, bee colonies, and human neuron system. Artificial Intelligence (AI) is a concept of study and research for finding relationships between cognitive science and computation theories [1]. These relationships are represented as data structures, search techniques, problem solving methods, or representation forms for knowledge [1]. These systems are working either under supervision or without supervision. Most of social insects work without supervision. In fact, their teamwork is largely self-organized, and coordination arises from the different interactions among individuals in the system [2]. These interactions might be primitive, like ants follow odor trails, or more complex, like a honey bee dancing. The collective behavior that emerges from a group of social insects has been artificially represented as a technique known as

Swarm Intelligence (SI) [3]. However, SI is a type of AI where the term swarm is used in a general manner to refer to any restrained collection of interacting agents or individuals [2]. SI systems are typically made up of a population of self-organized individuals interacting locally with one another and with their environment. Although there is normally no centralized control structure dictating how each individual should behave, local interactions between all individuals often lead to the emergence of global behavior. The rest of the paper is organized as follows. Section 2 looks at the related work and background of the algorithms used. Section 3 presents the algorithm developed in this research. Finally, Section 4 provides conclusions and offers avenues for future work. 2 BEES BEHAVIOR BACKGROUND

A bee colony can be thought of as a swarm whose individual agents are bees. Each bee at the low-level component works through a swarm at the global level of component to form a system. Thus, the system global behavior is determined from it is individual's local behavior where the different interactions and coordination among individuals lead

to an organized teamwork system. This system is characterized by the interacting collective behavior through labor division, distributed simultaneous task performance, specialized individuals, and selforganization. The exchange of information among bees leads to the formation of a tuned collective knowledge. A colony of honey bees consists of a queen, many drones (males) and thousands of workers (nonreproductive females). The queen's job is to lay eggs and to start new colonies. The sole function of the drones is to mate with the queen and during the fall they are ejected from the colony. The worker bees build honeycomb, and the young, clean the colony, feed the queen and drones, guard the colony, and collect food. As nectar is the bees' energy source, two kinds of worker bees are responsible for food. These are scout bees and forager bees. A bee does many things in its life history, and does not become a scout/work bee until late in its life [4]. While scout bees carry out the exploration process of the search space, forager bees control the exploitation process. However, exploration and exploitation processes must be carried out together by the colonys explorers and colonys exploiters. As the increase in the number of scouts encourages the exploration process, the increase of foragers encourages the exploitation process. Studying the foraging behavior leads to optimal foraging theory that directs activities towards achieving goals [5]. This theory states that organisms forage in such a way as to maximize their intake energy per unit time [5]. In other words, the swarm of bees behaves in such a way as to find and capture the food that containing the most energy while expending the least possible amount of time in real variables. There are two forms of scenarios for any bee in forging process which are either scout or forager. The following subsections present these two scenarios: 2.1 The Behavior of Scout Scenario Scouts fly around and search for food. When they find a source of nectar or pollen fly back to the colony and start dancing to communicate with other bees on a particular region in the comb. Fig. 1 presents the decoding of the language of the Bee dance [4]. Hence the behavior of the scout scenario is summarized according to the following activities: The scout flies from its colony searching for food sources in a random way. Once it finishes a full trip, it returns back to its colony. When a scout arrives at the colony, it goes inside and announces its presence by the wing vibrations. This means that it has a message to communicate. If it has found a nearby source of nectar or pollen,

it undergoes a circular dance. The nearby bees follow it through this circular dance and smell it for the identity of the flowers. They listen to the intensity of the wing vibrations to indicate the value of the food source. If the source is so close, no directions are given. Alternatively, if the flower source is a long ways off, careful directions must be given. The abstract convention that the scout makes is that the up position on the comb is the position of the sun. Because bees can see polarized light, they can tell sun position without actually seeing the sun. The scout dances in a precise angle from the vertical. This equals to the horizontal angle of the sun with reference to the colony exit with the location of the food source. Next the scout bee must tell the other bees how far away the flower source is. This is done by waggling the abdomen from side to side. The slower the waggling, the further away is the distance of the food flower.

Figure 1: Decoding the language of the Bee dance Thus the dance of scouts points to the direction, distance, and quality of food source. What Von Frisch notes is that the various groups of scouting bees compete with each other and that therefore the decision is finally made in favor of the best domicile [4]. 2.2 The Behavior of Forger Scenario The reaction of the forager bees on this show concludes into steps: The bees in the colony closely follow the scout to learn these directions, and also learn the odor of the flower on scout bee, so they can find the flower when they arrive at the source location. Because the sun is moving in the sky, the bees should use an accurate clock sense to adjust for the

changing sun position with reference to the food source and the colony exit. Even more remarkable, if a trained bee is removed from the colony to another location where the flower is not visible, but the colony is, the bee does not return to the colony to get its bearing, but reads sun position, and triangulates, and flies directly to the flower [4]. Subsequently, the forager bees take a load of nectar from the source and return to the colony and unload the nectar to the store of food. Foraging requires energy and the honeybees evaluation as to where, what, and how long to forage are all related to the economics of energy consumption and the net gain of food to the colony. Generally bees fly only as far as necessary to secure an acceptable food source from which there is a net-

gain. Therefore, these are the factors that influence foraging behavior and determine profitability. The net rate of energy intake is defined as the energy gained while foraging minus the energy spent on foraging, divided by time spent foraging. [6] 3 PROPOSED ALGORITHM

A modeling of artificial bee colony system has been proposed in [7], as seen in Fig. 2. This paper proposes a new optimization algorithm that uses the bee behavior in food forging as the functions to be used by the processing engine. Fig. 3 illustrates a simple example of Bees-Colony area with four orbits and six food sources: FS1,1 , FS3,1 , FS3,2 , FS3,3 , FS4,1 , and FS4,2.

Interacting with Environment

Functions

Processing Engine

Short Term Memory

Long Term Memory Figure 2: Architecture of Artificial Bees Colony System

FS3,3

Orbit 1 Orbit 2

FS4,2
Colony

Orbit 3 Orbit 4

FS3,2
FS3,1 FS4,1

FS1,1

Figure 3: Bees-Colony Area

The following notations are used in the proposed algorithm: 1. X: number of scouts. Sx: the xth scout bee. 2. Y: number of foragers. Where FRy means the yth forager bee. 3. N: number of bees in the colony population. Where, N = X + Y. Moreover, X = from 5% to 10% of N 4. I: number of orbits. Where i stands for the ith orbit. 5. CM: colony moral, it is a parameter defined algorithm goals. 6. FSij: Which means the jth food source at the ith orbit? 7. M: Number of maximum iterations. 8. FSQij: Means the quality of the jth food source at the ith orbit. 9. Dij: the direct distance between colony and FSij (see Fig. 3). 10. FSGij: the net gain of energy from the food source, FSij . It is calculated based on the distance, Dij , and the food source quality, FSQij. Where: FSGij = FSQij / Dij Table 1: Food Taboo Table (FTT) FS FS11 FS12 FS21 FS22 FSij Dij FSQij SRx

11. FSQTij: food source total quality is calculated based on food source net gain, FSGij, and the scout's rank SRx which find this food source in this case the scout with a better food source will be rewarded. Where FSQTij = FSGij + SRx. 12. SRx: the rank of the xth scout and this value will be updated every time forger chooses the source found by this scout. 13. FSVij: number of visits by the forgers for FSij. There are two tables that will be used by the proposed algorithm, these are: FTT: Food Taboo Table, as seen in Table 1, this table contains all food sources that had been visited by the scouts. It is used to prevent more scouts from visiting the same food source. This table contains the parameters: FSij, Sx and Srx. JST: Job Sheet Table, as seen in Table 2, any food source visited by scouts will be added to this table. Each forager picks up one of the food sources, FSij, to collect food from. This table contains the following.

FSGij

FSQTij

FSVij

Table 2: Job Sheet Table (JST) FS FS11 FS12 FS21 FS22 FSij Sx Srx

The following algorithm represents the main steps of the proposed algorithm that used in this paper: 1) Initialization: a) X: number of scouts. b) Y: number of foragers. c) N: number of bees in the colony population. d) I: number of orbits. e) CM: colony moral. f) FSij: the jth food source at the ith orbit. g) M: number of maximum iterations. h) FSQij: the quality of the food source, FSij . i) Dij: the direct distance between colony and FSij (see Fig. 3). 2) Assumptions: a) The food sources have been previously defined. b) Every scout bee, Sx , visit only one food source, FSij , per each trip. 3) REPEAT X times (i.e. all scouts, X, finish one trip): { a) Scout should choose one food source, FSij, find its distance, Dij , quality FSQij, and net gain of energy FSGij: FSGij = FSQij / Dij b) Update JST table. c) Update FTT table. } REPEAT UNTIL the CM goal is achieved or reached the maximum number of iterations, M: { REPEAT Y times (i.e. all foragers, Y, complete one trip): { (a) Generate a random number, RN. (b) Calculate food source total quality FSQTij = FSGij + SRx (c) Foragers, FRy , choose which food source, FSij , to forage by comparing the random number, RN, with FSQTij on the JS table. If FSQTij > RN Then Choose FSij and go to (h) Else Repeat (g) End If (d) Update JST. (e) Update the CM. } }

algorithm can be applied to many combinatorial optimization problems, dynamic problems in real variables, stochastic problems, multi-targets; data mining search engine crawling, parallel implementations, and any area not required supervision. 5 REFERENCES

[1] Owaied H. Hussein , Abu-Arra M. Mahmoud :Functional Model of Human System as knowledge Base System , The International Conference on Information & Knowledge Engineering , pp. 158-161 (June 2007). [2] Swarm Intelligence: A Whole New Way to Think About Business, Eric Bonabeau and Christopher Meyer, Harvard business review, r0105g, (May 2001). [3] Beni, G., and J. Wang. Swarm Intelligence. In Proceedings Seventh Annual Meeting of the Robotics Society of Japan, Tokyo, RSJ Press, pp. 425-428 (1989). [4] Von Frisch, Karl: Decoding the Language of the Bee, Science, Volume 185, Issue 4152, pp. 663668 (1974). [5] Kamil, Alan C., John R. Krebs and H. Ronald Pulliam, Foraging Behavior, Plenum Press, New York and London (1987). [6] Corbet, S.A., Kerslake, C.J.C., Brown, D. & Morland, N.E. Can bees select nectar-rich flowers in a patch? Journal of Apicultural Research, 23, pp. 234242 (1984). [7] Hussein Owaied, Saif Saab: Modeling Artificial Bees Colony System. IC-AI: pp. 443-446 (2008). [8] T. D. Seeley, The Wisdom of the Hive (Harvard University Press, Cambridge, MA,(1995).

CONCLUSION

This paper proposes an algorithm for the bee colony system according to the two scenarios of scouting and forging processes. The proposed