Available online at www.sciencedirect.

com

Expert Systems with Applications

Expert Systems with Applications 36 (2009) 30343047 www.elsevier.com/locate/eswa

An attribute-based ant colony system for adaptive learning object recommendation

Yao Jung Yang a,b,*, Chuni Wu a
a

Department of Information Management, Hsing-Kuo University of Management, No. 89, Yuying Street, Tainan 709, Taiwan b Department of Information Technology, Soochow University, No. 1, Shizi Street, Jiangsu, China

Abstract Teachers usually have a personal understanding of what good teaching means, and as a result of their experience and educationally related domain knowledge, many of them create learning objects (LO) and put them on the web for study use. In fact, most students cannot nd the most suitable LO (e.g. learning materials, learning assets, or learning packages) from webs. Consequently, many researchers have focused on developing e-learning systems with personalized learning mechanisms to assist on-line web-based learning and to adaptively provide learning paths. However, although most personalized learning mechanism systems neglect to consider the relationship between learner attributes (e.g. learning style, domain knowledge) and LOs attributes. Thus, it is not easy for a learner to nd an adaptive learning object that reects his own attributes in relationship to learning object attributes. Therefore, in this paper, based on an ant colony optimization (ACO) algorithm, we proposed an attributes-based ant colony system (AACS) to help learners nd an adaptive learning object more eectively. Our paper makes three critical contributions: (1) It presents an attribute-based search mechanism to nd adaptive learning objects eectively; (2) An attributes-ant algorithm was proposed; (3) An adaptive learning rule was developed to identify how learners with dierent attributes may locate learning objects which have a higher probability of being useful and suitable; (4) A web-based learning portal was created for learners to nd the learning objects more eectively. 2008 Elsevier Ltd. All rights reserved.
Keywords: Adaptive learning; Ant colony optimization; Learning style

1. Introduction In the classic teacher-centered situation, the course is built from the content dened by the teacher or author, and most teachers or educators agree that, in the design and development of educational material, attention must be focused on learner characteristics and requirements and dened in terms of content and learning style. Teachers usually have a personal understanding of what good teaching means as a result of their experience and educational related domain knowledge level in education, and
Corresponding author. Address: Department of Information Management, Hsing-Kuo University of Management, No. 89, Yuying Street, Tainan 709, Taiwan. Tel.: +886 6 2871511; fax: +886 6 2870917. E-mail addresses: danny.yang@mail.hku.edu.tw (Y.J. Yang), wu. chuni01@msa.hinet.net (C. Wu). 0957-4174/$ - see front matter 2008 Elsevier Ltd. All rights reserved. doi:10.1016/j.eswa.2008.01.066
*

they create learning resources to put on the web for the purpose of study. In fact, most students cannot nd the most suitable learning objects from the web because each LO has dierent attributes (e.g. learning object level, learning type), and each individual learner also has dierent characteristics or attributes (e.g. learning style, domain knowledge level). Adaptive learning provides an alternative to the traditional one size ts all approach and has driven the development of teaching and learning towards a dynamic learning environment. Thus, getting an adaptive LO to suit learners personalized needs is an important issue. Two major problems arise here: the one size t all approach gives the same learning materials to each learner (Brusilovsky, 2001; Stewart, Cristea, Brailsford, & Ashman, 2005), and the immense amount of information available leads to information overload (Berghel, 1997). Thus, adaptive learning has gained more attention in recent

Y.J. Yang, C. Wu / Expert Systems with Applications 36 (2009) 30343047

3035

years (Manouselis & Sampson, 2003; Melis et al., 2001; Weber, 1999). Several intelligent proposals have been developed: dynamic learning recommendation (Chen, Huang, & Chu, 2005; Huang, Chen, Huang, Jeng, & Kuo, 2007), intelligent learning contents suggestions (Huang, Chen, Kuo, & Jeng, 2007), adaptive pedagogical path (Semet, Jamont, Biojout, Lutton, & Collet, 2003; Semet, Lutton, & Collet, 2003), and adaptive learning rrez, 2007; Wang, (Canales, Pen a, Peredo, Sossa, & Gutie Wang, & Huang, 2007). Establishing the learning path of learner is certainly not a new approach as indicated above; but learner characteristics and the learning behaviors of learners have resulted in the development of an adaptive system. In Fig. 1, a scenario shows that a learner named Daniel wants to get an SQL relative learning object. Which LO is the best choice? To determine this, we must consider two core factors: the learners attributes and the attributes of the learning object. Each learner has dierent attributes, such as learning style and domain knowledge level (e.g. apprentice, beginner, intermediate, expert); each LO has its own attributes also, such as dierent types (e.g. text, video, etc.) and dierent levels (e.g. introductory, professional). Kolb (1974), Felder and Silverman (1988) and Larkin-Hein and Bundy (2001) indicated that students learn in many dierent ways. Some learn by seeing and hearing, others by feeling and doing; some focus on acting. Thus, learning styles should be considered as in order to develop a dynamic adaptive learning environment (Kettel, Thomson, & Greer, 2000; Robert, Lorne, & Maung, 1990). In the scenario, matching the learners requirements according his attributes, in order to assist him in nding the adaptive learning objects eciently is a signicant problem. In the last decade, swarm intelligence, which can provide an impressive level of adaptability for learners in this sort of dynamic learning environment has been reported to have

applications in many elds. Ant colony system (ACS; see Dorigo & Gambardella, 1997) aim at exploring an optimal learning object and predicating several possible pedagogical items. Employing an extended ant colony system approach based on Kolbs learning style model, the attribute-based ant colony system (AACS) is proposed to construct a search mechanism for nding a suitable learning object. AACS is dierent from many existing user models because it relies on the attributes of the learning objects of each learner in order to nd the optimal learning object. The remainder of this paper is organized as follows: Section 2: introduces Kolbs learning style model and ant colony system and optimization algorithm. Section 3: proposes an attributes-based ant colony system (AACS) for an adaptive learning object search mechanism. Section 4: makes experimental evaluations for parameters tune and heuristic information. Section 5: creates an application for AACS and a web portal for the adaptive learning object storage and retrieval. Section 6: conclusion. 2. Relative work In the last decade, the idea of the learning object (LO), which can simplify and make reusable a large package of learning resources has emerged. IEEE, the IMS Global Learning Consortium, the advanced distributed learning (ADL) co-laboratory, and others (MASIE Center, 2002) have produced learning objects standardized works. Currently, the international specication, Sharable Contents Objects References Model (SCORM) (ADL, 2004), that is based on the results of work done by the above mentioned groups, is widely used in the e-learning ecosystem. Thus, these related learning objects can be reorganized into a course sequence (ADL, 2004; IMS, 2003). Accordingly, many web-based tutoring systems have developed sophisticated solutions for customizing learners learning needs to

Fig. 1. Which one has the most adaptive LOs for the learner?

3036

Y.J. Yang, C. Wu / Expert Systems with Applications 36 (2009) 30343047

Fig. 2. Kolbs learning styles.

yield meaningful learning relationships. Course maintenance systems, adaptive courseware generations, and dynamic courseware generations (Brusilovsky & Vassileva, 2003) are the core approaches developed in this eld. The afore-mentioned approaches are based on using learning activities and education items to predict optimal paths. Thus, if this model can be associated with a sophisticated adaptation technique, such as ant colony optimization (ACO), it will provide an intelligent analysis of the solution-well-trodden learning objects for learners. 2.1. Kolbs learning style model David Kolb published his learning styles model in 1984. The model gave rise to related terms such as Kolbs experiential learning theory (ELT), and Kolbs learning styles inventory (LSI). Kolbs learning theory sets out four distinct learning styles (or preferences), which are based on a four-stage learning cycle. As Fig. 2 shows, Kolb explains that dierent people naturally prefer a certain individual learning style. Here are brief descriptions of the four Kolb learning styles. 2.1.1. Diverging (feeling and watching CE/RO) These people are able to look at things from dierent perspectives. They are sensitive. They prefer to watch rather than do, tending to gather information and use imagination to solve problems. The learners who are categorized as Diverging style learners prefer learning objects in the form of animation, charts, graphs, ow charts or symbols.

2.1.2. Assimilating (watching and thinking AC/RO) The Assimilating learning preference is for a concise, logical approach. Ideas and concepts are more important than people. These people require good clear explanations rather than practical opportunity. The learners who are categorized as Assimilating style learner prefer learning objects in the form of the audio, video, lectures, verbal tutorials, etc. 2.1.3. Converging (doing and thinking AC/AE) People with a Converging learning style can solve problems and will use their learning to nd solutions to practical concerns. The learners who are categorized as Converging style learners prefer learning objects in the form of text-based materials, such as Microsoft oce (PowerPoint, Word, Excel), web pages, etc. 2.1.4. Accommodating (doing and feeling CE/AE) People with an Accommodating learning style will tend to rely on others for information rather than carry out their own analysis. The learners who are categorized as Accommodating style learner prefer learning objects in experience shared and practice activities (for example, tutorials on the web, web pages, SCORM1, LOM2, etc.) rather than in the form of animation, charts, graphs, ow charts or symbols.

SCORM (2004). Advance Distributed Learning Network Shareable Content Object Reference Model, retrieved October 10, 2004 from http:// www.adlnet.org/ 2 LOM (2004).IEEE WG 12: Learning Object Metadata, retrieved October 10, 2004 from http://ltsc.ieee.org/wg12/index.html

Y.J. Yang, C. Wu / Expert Systems with Applications 36 (2009) 30343047 Table 1 ACO algorithm according to chronological order of appearance ACO algorithm Ant system (AS) Elitist AS ANT-Q Ant colony system MaxMin AS Rank-based AS ANTS Hyper-cube AS TSP Yes Yes Yes Yes Yes Yes No No Main references Dorigo (1992) and Dorigo, Maniezzo, & Colorni (1991a,b, 1996) Dorigo (1992) and Dorigo, Maniezzo, & Colorni (1991a,b, 1996) Gambardella & Dorigo (1995) and Dorigo & Gambardella (1996) Dorigo & Gambardella (1997a,b) Stu tzle & Hoos (1996, 2000) and Stu tzle (1999) Bullnheimer, Hartl, and Strauss (1997, 1999) Maniezzo (1999) Blum, Roil, & Dorigo (2001) and Blum & Dorigo (2004)

3037

probability that other ants will use that path, and therefore the pheromone trail on such a path will grow faster and attract more ants to follow (so called positive feedback). An iterative local search algorithm tries to search the current paths to neighboring paths until a better solution is found. Informally, an ACO algorithm can be imagined as the following algorithm:

In the column TSP we indicate whether this ACO algorithm has already been applied to the traveling salesman problem.

procedure ACO algorithm Set parameters, initialize pheromone trails while (termination condition not met) do Construct Solution Apply Local Search Update Pheromone Trails End end

2.2. Ant colony optimization (ACO) algorithm for nding adaptive Learning Objects Ant colony optimization is a metaheuristic in which a colony of articial ants cooperates in nding good solutions to dicult optimization problems. A metaheuistic is a set of algorithm concepts that can be used to dene heuristic methods applicable to a wide set of dierent problems. The use of metaheuristics has signicantly increased the possibility of nding high quality solutions to hard, practically relevant combinatorial optimization problems within a reasonable time. The traveling salesman problem (TSP) plays an import role in ACO research: the rst ACO algorithm, call ant system, proposed by Dorigo, was successfully applied in tacking the well-known TSP (Dorigo & Gambardella, 1997), as well as many of the ACO algorithms proposed subsequently, in shown in Table 1 (Dorigo & Stu tzle, 2004). The ant colony system (ACS) is a particular algorithm of ACO which is based on agents that simulate the natural behavior of ants, develop mechanisms for cooperation, and assist them in using experience (Dorigo & Gambardella, 1997) to nd the shortest path between a food source and the nest. ACS is a population-based heuristics that enables the exploration of the positive feedback whereas the ants are able to communicate (ants lay pheromone for indirect communication, so called stimergy3) information concerning food source via an aromatic essence. The ants lay pheromone and heuristic information to mark trails. As the paths are visited by other ants, some of the trails may be reinforced and others paths may be allowed to evaporate. Pheromone trails can be observed via the number of ants passing through the trail. When there are more pheromones on a path, there is larger

Here are brief descriptions of the ACO algorithm. 2.2.1. Construction Solution The problem can be represented in a nite set of components C, set C fc1 ; c2 ; . . . ; cN c g, where Nc is the number of the components, and the states of the problem are dened in term of sequence x = hci, cj, . . ., ch, . . .i of nite length over the elements of C. The set of all possible states is denoted by X. The set of solution S is subset of X (i.e., S # X). Given this formulation, the ants build solutions by performing randomized walks on the completely connected graph GC = (C, L) whose nodes are the components C, and the set L fully connects the components C. the GC is called the construction graph and the elements of L are called the connections. At each construction step, ant k applies a probabilistic action choice rule, called random proportional rule, to decide which node to visit next. In particular, the transition probability with which ant k, currently at node i, chooses to go to node j is sij ta gij tb ; Pk ij t P sil ta gil tb
l2N k i

8j 2 N k i; k 2 1 $ m

2:1

, 1959; p. 79). The original denition of stimergy (see Grasse

where gij = 1/dij is a heuristic value that is available a priori, a and b are two parameters which determine the relative inuence of the pheromone trails and the heuristic information, and N k i is the feasible neighborhood of ant k when located at node i and the set of nodes that ant k has not visited yet (the probability of choosing a node outside N k i is 0). By this probabilistic rule, the probability of choosing a particular arc (i, j) increases with the value of the associated pheromone trails sij and according to the heuristic information value gij. The role of parameters a and b determine the consideration of pheromone trails and heuristic bias.

3038

Y.J. Yang, C. Wu / Expert Systems with Applications 36 (2009) 30343047

2.2.2. Update of Pheromone trails After all the ants have constructed their tours, the pheromone trails are updated. This is done by rst lowering the pheromone value on all arcs by a constant factor, and then adding pheromone on the arcs the ants have crossed in their tours. Pheromone evaporation is implemented by sij 1 qsij ; 8i; j 2 L; 2:2

where 0 < q 6 1 is the pheromone evaporation rate. After evaporation, all ants deposit pheromone on the arcs they have crossed in their tour: m X sij Dsk 8i; j 2 L; 2:3 sij ij ;
k 1

is the amount of the pheromone ant k deposits where on the arcs it has visited. It is dened as follows: ( 1=C k ; if arci; j belongs to T k ; D sk 2:4 ij 0; otherwise; where Ck, the length of the tour Tk build by the kth ant, is computed as the sum of the lengths of the arcs belonging to Tk. 2.2.3. Heuristic information It is represented as a consciousness constant, which is the heuristic preference of moving from nodei to the next node j. It is set to 1/dij, where dij is the distance between node i and node j. 2.2.4. Local search Pheromone evaporation and pheromone deposits take place only on the arcs belonging to the best-so-far tour, when ant located at node i, ant k, moves to node j chosen according to the so-called pseudorandom proportional rule, given by q n o ( b s ; if q 6 q0 ; argmaxh2N k ih gih i 2:5 j J otherwise; where q is a random variable uniformly distributed in [0, 1], q0 (0 6 q0 6 1) is a parameter, and j is a random variable selected according to the probability distribution given by Eq. (2.1) (with a = 1). In other words, with probability q0 the ant makes the best possible move as indicated by the learned pheromone trails and heuristic information (the ant is exploiting the learned knowledge), while with probability (1 q0) it performs a biased exploration of the arcs. Thus, the update in ACS is implemented by the following equation: sij 1 qsij qDsbs ij ; 8i; j 2 T
bs

Dsk ij

The Elitist Ant System (EAS) was introduced in Dorigo (1992) and Dorigo et al. (1991a, 1996). The idea is to provide strong additional reinforcement to the arcs belonging to the best tour found since the start of the algorithm; note that this additional feedback can be viewed as additional pheromone deposited by an additional ant called best-sofar ant. A Rank-Based Ant System was proposed by Bullnheimer et al. (1999). In Rank-Based Ant System, each ant deposits with its rank. Cordon, Herrera, Fernandez de Viana, and Moreno (2000) used the transition rule and pheromone evaporation mechanism to improve the ants solutions. Most recently, Semet, Lutton, et al. (2003) and Semet, Jamont, et al. (2003) applied the ACO heuristics to an e-learning pedagogic material navigation problem, and Jamont, Collect, Lutton, Biojout, and Bourgeois Republique (2005) experimented with an ant-hill method which laid the pheromone depending on how students validated an item, so as to optimize learning paths with dierent students who have dierent views. Thus, the ACO method seems well suited for tackling adaptive learning object searches in a dynamic learning environment. 3. Attribute-based ant colony system (AACS) In this section, we present attribute-based ant colony system (AACS) as a method of nding a learning object and then provide recommendations for the adaptive learning object for learners. AACS is derived from an extension of the ant colony system that updates the trails pheromones from dierent knowledge levels and dierent styles of a groups learners to create a powerful and dynamic learning object search mechanism. In order to achieve this, there are three prerequisites to applying AACS, that (a) the adaptive learning portal knows the learners attributes, and the attributes include the learners knowledge level and learning style (the learners have been referred to previously using Kolbs learning style). (b) The learners attributes (include Kolbs learning style and learners knowledge level) and LOs attributes (learning object type and learning object level) have been annotated by teachers or content providers. For example, learners who are categorized as Diverging style learners prefer learning objects in the form of audio or video (i.e., audio, video, animation), rather than Text-style learning objects. The AACS is further used as a prerequisite to match the relationships between learners and the learning object. 3.1. Learners activities and attributes recorded into the Learning Object Repository In FORPA4 (Yang & Song, 2006), the learning portal can gather the learning activities of each learner in an online course and then record their paths and attributes into
4 FORPA is an adaptive learning portal which is a real instance for AACS mechanism.

2:6

bs bs bs where Dsbs ij 1=C , which C is the length of the tour T . It is important to note that in the pheromone trails update, both evaporation and new pheromone deposit, only applies to the arcs of Tbs and not to all the arcs. Many reasearchers have given ACO their attention to develop sophisticated models and to extend its algorithm.

Y.J. Yang, C. Wu / Expert Systems with Applications 36 (2009) 30343047

3039

a learner transaction database (Learning Object Repository). Given the denition of a learning object relationship, let R = {r1, r2, . . ., rn}, R is indicated as a nite set of learning objects relationships, which is expressed as the following traversal sequence of a learner. For example: R = {ra, rb, rc, rd}, (learning objects {a, b, c, d} to simplify), and we have paired these objects as transaction data (a database record) that is veried as a forward sequence of {(a, b), (a, c), (c, d), (d, a), (d, b), (b, c)}, where pair (a, b) denotes the relationship between the learning objects a and b, and the learning object b is the next learning object from a. Furthermore, the transaction data of the learners and learning object relationship can be formally stored as a transition matrix in a learning object repository. 3.2. The Attribute ants and pheromone update rule Since learning objects may be continuously added into object repositories by learning content providers at any time, each learner has his own attributes, and the learning object also has its own attributes, as shown in Table 2, we propose the idea of an attribute ant to make it easy for a learner to get the most suitable learning object based on his own attributes and LOs attributes. The attributes ant mainly uses an extended learner model and object level to involve the adaptive rule in a dynamic learning setting. Following this rule, the system can improve the quality of the pheromone as well as help learners easily nd their own adaptive learning object. In AACS, we dene two rules to reinforce pheromone update as following: Rule 1: IF learners attributes MATCH the LOs attributes THEN Daemon action with MRstyle and MRlevel. Rule 2: IF learners attributes PARTIALLY MATCH the LOs attributes THEN Daemon action with MRstyle or MRlevel. There is a formal denition of a MATCH and PARTIALLY MATCH: 1. IF learner learning style = learning object type AND Domain Knowledge Level = learning object levelTHENwe call it a MATCH. 2. IF (learner learning style = learning object type AND Domain Knowledge Level h i learning object level) OR (learner learning style h i learning object type AND Domain Knowledge Level = learning object level) THEN we call it a PARTIALLY MATCH. In AACS, whether or not the heuristic information can be reinforced relies on the attribute ant. We give a formal denition of attribute ant: let k ant, which has one of the Kolbs learning styles and its own domain knowledge level, generate a xed amount of pheromone when it travels along a node. If the k ants attributes MATCH or are a PARTIALLY MATCH with the node attributes

Table 2 The learners attributes and LOs attributes Learner attributes Kolbs learning style 1.Diverging 2.Assimilating 3.Converging Knowledge level 1.Apprentice 2.Beginner 3.Intermediate Learning object attributes Learning object type 1.Graphic(image, charts, symbol), 2.Video(audio, animation) 3.Text(word, PowerPoint, Excel) 4.XML(Web, SCORM, LOM) Learning object level 1.Initial 2.Introductory 3.Advance

4.Accommodating

4.Expert

4.Professional

(learning object attributes), then the nodes which are on the traveled path created by an ant k may obtain an extra pheromone after each time unit. In order to further explain the idea of the attribute ant, we illustrate using an example for explaining the adaptive rule. As show in Fig. 3, when a learner searches from the node a to its neighbor nodes, the next node is possibly b, c and d. However, involving the adaptive rule, the learner with an Assimilating learning style and his domain knowledge level marked Beginner, the node c with a Video type and object level mark Introductory. Thus, the intensity of node c has a higher probability of being recommended than node b and d for the learner with Assimilating style and Beginner level. 3.3. The proposed algorithm We propose the attribute-based ant colony system (AACS) is an extension of the ant colony system, and the parameters and functions used in this paper are the same as those dened in ACO. However, we applied two rules for pheromone update and heuristic information. 3.3.1. The heuristic information The heuristic value gij is a normalized value function of the queue length qij (the learner waiting to be processed) on the node connecting the node i with its neighbor j The heuristic information is dened as the following: , ! Ni X qil 3:1 gij 1 qij MRij
l1

gij gives a quantitative measure associated with the node waiting time and match ratio (MR, given by Eq. (3.7)). An ants decisions are therefore taken on the basis of the combination of long-term learning process and an instantaneous heuristic prediction, and it is implied in the denition as the required capability of a learner to achieve the next learning objective. In other words, a higher value of gij means there is a higher probability of being able to choose the high match ratio nodes, and to get the approximately optimal adaptive LOs.

3040

Y.J. Yang, C. Wu / Expert Systems with Applications 36 (2009) 30343047

Fig. 3. A MATCH, PARTIALLY MATCH scenario.

3.3.2. Pheromone trail update The relational strength between the ith node and the jth node is pheromone trail intensity sij. The incremental intensity Dsij(t) is that which locates pheromone trail value at the time t, and is updated as the following formula: sij t qsij t 1 Dsij t 3:3

MRstyle eMR ij MRlevel ij e

Styleij

3:8 3:9

MR Levelij

where S is the learner style, T is the learning object type, K is the learner knowledge level, and L is the learning object level. 3.3.3. Transition probability of attributes ants The probability P k ij t given in Eq. (2.1) is dened as a compromise value of sij and gij where the k ant move forward from a learning object i to the next object j, j is one of the neighbors of i, j 2Nk(i). The role of parameter a and b is the following: If a = 0, this corresponds to a classic stochastic greedy algorithm. If b = 0, only pheromone amplication is at work, that is, only pheromone is used, without any heuristic bias. This generally leads to poor results, in particular, for values of a > 1, it leads to the rapid emergence of a stagnation situation, that is, a situation in which all the ants follow the same path and construct the same tour, which, in general, is strongly suboptimal (Dorigo, 1992: Dorigo et al., 1996). 3.3.4. The adaptive procedure for AACS algorithm We proposed that the AACS algorithm has ve main procedures for resolving the problem of adaptive paths, and pseudo-code of the attribute-based ant colony system is shown as the following:

where q is the evaporate ratio of the trail at an interval time unit. If an attribute ant has chosen the jth node after locating node i and laid its pheromone trail, the pheromone levels on the jth node should be updated and the contributions of all attribute ants, and the amounts of the pheromone laid by the ants is dened as below,
level sij t qsij t 1 w Dsstyle ij t Dsij t

3:4

level where Dsstyle ij t and Dsij t are the variable amounts of pheromone deposited in the arc(i, j), and represent those learner who MATCH their attributes (attribute ants) with learning object attributes. Additionally, we give a weighted variable w to adjust the match ratio of the learning style factor. Thus, the number of attributes ants has to be considered; the adaptive solution is the following: k ;style t Dsij m X m n Q MRstyle ij t n0 k ;style t on the arc(i, j) where m is the number of MRij k ;level Dsij t m X n0

3:5

m n Q MRlevel ij t

3:6

where c is a constant number to adjust actual learning obk ;style t on the ject search situations, m is the number of MRij style arc(i, j), MR is the match ratio for learner learning style, and MRlevel is the match ratio for learning object level, the denition of match ratio given as follows: X X X MRij MRstyle MRlevel 3:7 Let ij ij
8j2N i 8j2N i 8j2N i

where MR Styleij S i T j = 8r2N i S i T r 1 X K i Lr 1; MR Levelij K i Lj =

8r2N i

and

procedure Attribute_ACS/* main procedure*/ Initialize_Parameter while (condition not terminate) do Construct_Solution Heuristic_Decision_Rule Update_Pheromone_Trails if (Match the adaptive rule) then Daemon_Actions end while Recommended_Learning_Object end procedure

Y.J. Yang, C. Wu / Expert Systems with Applications 36 (2009) 30343047

3041

procedure Initial_Parameter set the Q for trail constant intensity, heuristic factor, a, b and evaporate rate q, and setting the pheromone trails to a initial value s0> 0; Initializes learner attribute; set diverging = 1, Assimilating = 2, converging = 3, accommodating = 4; set apprentice = 1, beginner = 2, intermediate = 3, expert = 4; initializes learning object attribute; set Graphic = 1, Video = 2, Text = 3, XML = 4; set initial = 1, introductory = 2, advance = 3, professional = 4; end Initial_Paramete

computing the PN iheuristic information gij 1 qij MRij = l 1 qil Pk ij t

sij ta gij tb

8j 2 N k i;k 2 1 $ m end foreach

l 2N k i

sil t gil t ;

end Heuristic_Decision_Rule

procedure Update_Pheromone_Trails set the adding pheromone with evaporate factor (1 q)*sij "(i, j) 2L where 0 < q 6 1, Dsij computing P all ants deposit pheromone sij m 8i; j 2 L sij k 1 Dsij ; end Update_Pheromone_Trails

procedure Construct_Solution select a start node x1 and put into construct node list xL while (xL R Sc and N k i / do j Select_Next_Node (xL, R) put j into xL; end while if xL 2 S then return xL else abort end-if end Construction_Solution

procedure Daemon_Actions foreach (attributes ant MATCH the adaptive rule) computing global pheromone update with attribute match ratio Pm ;style k ;style t n0 m n Q MRij t Dsk ij P m k ;level k ;level Dsij t n0 m n Q MRij t level sij t qsij t 1 Dsstyle ij t Dsij t end foreach end Daemon_Actions

procedure Heruistic_Decision_Rule let S = current_learner_learning_style for S = 1, . . . 4, let T = current_learning_object_type for T = 1, . . . 4, let K = current_learner_domain_knowledge_level for K = 1, . . . 4, let L = current_learning_object_level for L = 1, . . . 4, P Computing MR Styleij S i T j = 8r2N i S i T r 1 P MR Levelij K i Lj = 8r2 N i K i L r 1 , MR Style eij , MRstyle ij level MR Level MRij eij P P style Let 8j2N i MRij w 8j2N i MRij P level 8j2N i MRij /* where w is a parameter for enhance learning style weighted, w > 1 */ foreach xij do get candidate j from N k i

procedure Recommended_Learning_Object List the learning object from node list xL Return adaptive learning object list end Recommended_Learning_Object

Below are brief descriptions of the main procedure function in AACS algorithm pseudo-code Initialize_Parameter: the procedure which initializes parameters of AACS, Q, a, b, q, learning style, domain knowledge level, object type and object level. Construct_Solution: the procedure for construct solution, where Sc is the set of candidate solutions, and N k i is set of the candidate nodes next to i for ant k, the random walk of ants is biased by pheromone trails which are associated with a connection between nodes i andj. The ant k selects the next nodes j from neighbor list based on the

3042

Y.J. Yang, C. Wu / Expert Systems with Applications 36 (2009) 30343047

Fig. 4. The result of (a) higher attribute-ant match ratio eciently causes daemon pheromone update. (b) Varying the number of ants used: the plot give the match ratio values deviation from the optimal tour as a function of the number of iterations for using a number of ants varying from 10 ants to 200 ants.

E-Education Portal

Isa

Kolb Learning Style

Diverging Assimilating
Isa Isa Isa Isa Login/Answer

Author Attribute
Login/Que stionnaire
Record

Accommodating Converging

Publish
Isa

Teacher
Isa Isa

Domain Expert Easy

Isa Isa Isa

Author Profile
Login/Register Isa Store/Reference

Isa

LR Objective

Advance

Facilitate Learning Style Beginner Intermidate

Isa Isa Isa Isa

Expert Easy
Isa Isa

Intermeate
Isa

Advance Expert
(HTML/XML) (WORD, PPT) (VIDEO/AUDIO) (SCORM,LMS)

Learner
Get

Learning Resource (LR)

Store

Isa

Resource Level

Web Page
Isa Isa Isa Isa Isa

Document Media Package

Expert
Isa

Domain Knowledge References Background

Cal l References

LRs Ontology Repository

Store

Resource Format/Ty Find pe

Learner Attribute
References

Resource Attribute
Ontology Search Agent

Fuzzy Rule Module

Justified

Learning LRs Objects Fuzzy-Ontology Repository Database

References

Ontology Search

Ontology Fuzzy A portal using AACS abase/ technique Rule&Search

Fig. 5. An adaptive portal (FORPA) architecture using AACS technique.

Y.J. Yang, C. Wu / Expert Systems with Applications 36 (2009) 30343047

3043

Fig. 6. A example for Oracle Database Courseware applied AACS system.

Select_Next_Node function, which mainly focused on the transition probability value P k ij , as shown in Eq. (2.1), after that, choose the next nodej from candidate nodes and put into optimal solution node list xL. Heruistic_Decision_Rule: the procedure which computes the heuristic information gij and match ratio, where MRstyle is the match ratio for learner learning style, and MRlevel is the match ratio for learning object level. Update_Pheromone_Trails: this procedure computes the general pheromone update sij after evaporation, where Dsij is the amount of pheromone ant k deposits on the arcs it has visited. Daemon_Actions: the procedure which computes the global pheromone update value Dsij, where Dsstyle ij t and t are the variable amounts of pheromone deposited Dslevel ij in the arc(i, j), and represent those learners who MATCH

their attributes (attribute ants) with learning object attributes. Recommended_Learning_Object: the procedure which gets the adaptive learning object from solution nodes list xL, and then returns to the system. 4. Experimental evaluations of trails intensities This section describes the simulation of tuning parameters in AACS. Pilot experiments were conducted to optimize the parameters in order to match the actual situation. We observe the variation of the intensity of two trails. As shown in Fig. 4, the blue5 curve which utilizes
5 For interpretation of color in Fig. 4, the reader is referred to the web version of this article.

3044

Y.J. Yang, C. Wu / Expert Systems with Applications 36 (2009) 30343047

Fig. 7. FOAPA portal gets the adaptive learning object according to the learners attributes.

traditional pheromone rule, does not consider aects from learning style, and adds the population to take variables of trail intensity s into account in each time t. Considering the solid green curve as shown in Fig. 4 that uses our proposed rule (in Section 3.3.2, Eqs. (3.3)(3.9)) and considering the learning style of the learner, we dynamically add to the population in which the learning object has the same style each time. It was found that this curve has exponential growth. That is, if a node (as a learning object) has more learners with the same style traveling along a path with AACS operating, the higher the match ratio (MB) will become, and the node is possibly a candidate for the best choice. In Fig. 4, the experiment shows that higher MB value caused the attribute-ant to get the adaptive learning object eciently, and the proposed rule to update the trail intensity has a signicant enough amount of pheromone to aect the probability distribution along the paths as opposed to the traditional pheromone update rule. 5. Application: a web portal for adaptive learning path recommendation The architecture of a web-base adaptive learning portal was shown in Fig. 5, The modules of the FOAPA are described below. 5.1. Login/Questionnaire Upon entering the Portal for the rst time, each learner/ author is prompted with a short questionnaire, in order to

determine his or her characteristics, learning style, and domain knowledge level. This prole is automatically updated, taking into account the learner interactions with the Portal. 5.2. Learners The learners need to be able to authenticate themselves via login module, as well as to dene, edit and save their prole for adaptive personalized services. Then, learners can, via the AACS agent and match rule module, retrieve LOs according to their specic requirements, learning styles and domain knowledge levels. 5.3. Authors The authors need to be able to publish or upload their LOs in a commonly accessible format, so that they can be eectively, eciently and adaptively searched and retrieved in the dierent contexts of use described in LOs attribute. Dierent LOs are described in metadata which are related to each resource and presented according to different learner proles, thus meeting the requirements for learners-adaptive personalized learning. 5.4. AACS agent The AACS Agent is responsible for compiling the query result from the learning objects repository, ranking them according to user preferences and match ratio, as supplied by the match rule module. In particular, all the learning

Y.J. Yang, C. Wu / Expert Systems with Applications 36 (2009) 30343047

3045

Fig. 8. A screenshot for learner get the recommendation learning object.

objects were using Dublin6 (2004) metadata tags the learning object, and it a standard ontology object presentation form. 5.5. Match rule module The match rule module is responsible for dening criteria for searching and ranking learning objects. Through the rules which are dened in Section 3.3.2, the attributes-ant can access the adaptive learning object via the AACS algorithm quickly and eciently. 5.6. Ontology object repository Before storing into the ontology object repository, all the learning objects must be tagged with Dublin (2004) metadata tags, details about which have been published as ANSI/NISO Standard Z39.85-2007. 5.7. Learning objects database We create table schema for storage of the learners attributes and LOs attribute which can be accessed via the AACS agent and match rule module. The AACS agent uses attribute-ant behavior to get the adaptive learning object
6 DCMI (2007). Dublin Core Metadata Initiative, retrieved July 2, 2007 from http://dublincore.org.

via the match rule, and then returns the optimal result for the learner. We provide the ORACLE OCP (oracle certicated professional) courseware to be a real instance for AACS system, the example was shown in Fig. 6. In this example, shows that a learner how to get the adaptive learning object using the AACS mechanism, the higher MB value will lead to the high probability to get the most suitable learning object. The FORPA system screenshot was shown in Figs. 7 and 8. 6. Conclusion This work has proposed an attribute-based ant colony system (AACS) and introduced an instance of an adaptive learning portal, named FORPA. It is now available at http://192.192.111.113. AACS oer ecient heuristic information and match rules for enhancing the discovery of an adaptive learning object. It also provides an idea, named attributes-ant, which combined the Kolbs learning style model and the learners domain knowledge level with learning objects attributes to provide an adaptive solution for learners. The FORPA, a web-based learning portal with the AACS application, has been developed by the authors to support the adaptive learning device. The key concept for developing an adaptive learning system comes primarily from recognizing the dierent attributes of dierent learners. From the computer-supported learning perspective, there are two critical points: how a learning system launches

3046

Y.J. Yang, C. Wu / Expert Systems with Applications 36 (2009) 30343047 Dorigo, M., Maniezzo, V., & Colorni, A. (1991a). Positive feedback as a search strategy. Technical report 91-016 revised, Dipartmento di Elettonica, Politecnico di Milano, Milan. Dorigo, M., Maniezzo, V., & Colorni, A. (1991b). The Ant System: An autocatalytic optimizing process. Technical report 91-016 revised, Dipartmento di Elettonica, Politecnico di Milano, Milan. Dorigo, M., Maniezzo, V., & Colorni, A. (1996). Ant System: Optimization by a colony of cooperating agents. IEEE Transactions on System, Man, and Cybernetics - Part B, 26(1), 2941. Dorigo, M., & Stu tzle, T. (2004). Ant colony optimization. Bradford Books (Hardcover-Jul 1, 2004). Felder, R. M., & Silverman, L. K. (1988). Learning and teaching styles in engineering education. Engineering Education, 78(7), 674681. Gambardella, L. M., & Dorigo, M. (1995). Ant-Q: A reinforcement learning approach to the traveling salesman problem. In A. Prieditis & S. Russell (Eds.), Proceedings of the Twelfth international Conference on Machine Learning (ML-95) (pp. 252260). Palo Alto, CA: Morgan Kaufmann. , P. P. (1959). La reconstruction du nid er les coordinations Grasse interindividulles chez Bellicositermesnatalensis er Cubitermes sp. La orie de la stigmergie: Essai dinterpre tation du comportement des the termites constructeurs. Insectes Sociaux, 6. Huang, Y. M., Chen, J. N., Huang, T. C., Jeng, Y. L., & Kuo, Y. H. (2007). Standardized course generation process using dynamic fuzzy Petri nets. Expert Systems with Applications, 34(1). Huang, Y. M., Chen, J. N., Kuo, Y. H., & Jeng, Y. L. (2007). An intelligent human-expert forum system based on fuzzy information retrieval technique. Expert Systems with Applications, 34(2). IMS Global Learning Consortium, Inc. (2003). IMS learning design best practice and implementation guide. Version 1.0 nal specication. <http://www.imsglobal.org/learningdesign/ldvlp0/imsld_bestvlp0. html> Accessed 10.05.06. Jamont, Y., Valigiani G., Collect, P., Lutton, E., Biojout, R., & Bourgeois Republique, C. (2005). Experimenting with a real-size man-hill to optimize pedagogical paths. In SAC05 proceedings (pp. 1317). Mars, Santa Fe, NM, USA. Kettel, L., Thomson, J., & Greer, J., (2000). Generating indivisualized hypermedia applications. In P. Brusilovski (Ed.), Proceddings of ITS2000 workshop on adaptive and intelligent web-based education systems (pp. 2836), Montreal. Kolb, D. A. (1974). Learning style inventory technical. Boston McBer & Co. Larkin-Hein, T., & Bundy, D. D. (2001). Research on learning style: Applications in the physics and engineering classroom. IEEE Transaction on Education, 44(3), 276281. Maniezzo, V. (1999). Exact and approximate nondeterministic tree-search procedures r the quadratic assignment problem. INFORMS Journal on Computing, 11(4), 358369. Manouselis, N., & Sampson, D. (2003). Agent-based e-learning course recommendation: Matching learner characteristics with content attributes. International Journal of Computers and Applications (IJCA), 25(1). MASIE Center (2002). Making sense of learning specications & standards: A decision markers guide to their adoption. Industry Report. <http://www.masie.com/standards/S3_Guide.pdf> Accessed 18.06.07. Melis, E., Andres, E., Bu denberder, J., Frishauf, A., Goguadse, G., Libbrecht, P., et al. (2001). ActiveMath: A generic and adaptive webbased learning environment. International Journal of Articial Intelligence in Education, 12(4), 385407. Robert, P. B., Lorne, O., & Maung, K. S. (1990). The importance of learning style in end-user training. MIS Quarterly, 14(1), 101119. Semet, Y., Jamont, Y., Biojout, R., Lutton, E., & Collet, P. (2003). Articial ant colony and e-learning: An optimization of pedagogical paths. HCI. Semet, Y., Lutton, E., & Collet, P. (2003). Ant colony optimization for elearning: Observing the emergence of the pedagogic suggestions. IEEE Swarm Intelligence Symposium, April (pp. 4652).

an adaptive learning object and how dierent attributes of learning objects can conform to learners attributes and as a consequence, lead learners down a well-trodden path. In addition, the attributes-ant colony system proposed in this study has been demonstrated as a successful solution for adaptive learning object search mechanisms in an e-learning domain. The ndings suggest that dierent object attributes and learners attributes have enhanced the opportunity for adaptive learning that leads to improved learning outcomes. In the future, we would like to apply the AACS to various domain applications, such as supply-chain systems, intelligent education systems, visitor tour paths, job schedule systems, and so on. References
ADL (2004). Advance distributed learning initiative. Shareable content object reference model (SCORM) 204 (3rd ed.). Overview. ADL Web Site <http://www.adlnet.gov/scorm> Accessed 18.06.07. Berghel, H. (1997). Cyberspace 2000: Dealing with information overload. Communications of the ACM, 40(2). Blum, C., & Dorigo, M. (2004). The hyper-cube framework for ant colony optimization. IEEE Transactions on Systems, Man, and CyberneticsPart B, to appear. Blum, C., Roil, A., & Dorigo, M. (2001). HC-ACO. The hyper-cube framework for ant colony optimization. In Proceedings of MIC2001 Metaheuristics International Conference (vol. 2 pp. 399403). Brusilovsky, P. (2001). Adaptive hypermedia. In Alfread Kobas (Ed.), User modeling and user adapted interaction. Ten year anniversary issue (Vol. 11, pp. 87110). Brusilovsky, P., & Vassileva, J. (2003). Course sequencing techniques for large-scale web-based education. International Journal of Continental Engineering Education and Lifelong Learning, 13(12), 7594. Bullnheimer, B., Hartl, R. F., & Strauss, C. (1997). A new rank based version of the ant system A computational study. Technical Report, University of Viena, Institute of Management Science. Bullnheimer, B., Hart1, R. F., & Strauss, C. (1999). A new rank-nased version of the ant system: A computational study. Central European Journal for Operations Research and Economics, 7(1), 2538. rrez, A. (2007). Canales, A., Pen a, A., Peredo, R., Sossa, H., & Gutie Adaptive and intelligent web based education system: Towards an integral architecture and framework. Expert Systems with Applications, 33, 10761089. Chen, J. N., Huang, Y. M., & Chu, W. C. (2005). Applying dynamic fuzzy Petri net to web learning system. Interactive Learning, 13(3), 159178. Cordon, O., Herrera, F., Fernandez de Viana, I., & Moreno, L. (2000). A new ACO model integrating evolutionary computation concepts: The best-worst ant system. In Proceedings of the 2nd international workshop on ant algorithms, Brussels, Belgium (pp. 2229). Dorigo, M. (1992). Optimization, learning and natural algorithms. PhD thesis, Dipartimento di Elettronica, Politecnico di Milano, Milan. Dorigo, M., & Gambardella, L. M. (1996). A study of some properties of Ant-Q. In H. Voigt, W. Ebeling, I. Rechenberg, & H. Schwefel (Eds.), Proceeding of PPSN-IV, Forth International Conference on Parallel Problem Solving from Nature. Lecture Notes in Computer Science (Vol. 1141, pp. 656665). Berlin: Springer-Verlag. Dorigo, M., & Gambardella, L. M. (1997). Ant colony system: A cooperative learning approach to the traveling salesman problem. IEEE Transaction on Evolutionary Computation, 1(1), 5366. Dorigo, M., & Gambardella, L. M. (1997a). Ant colonies for the traveling salesman problem. BioSystems, 43(2), 7381. Dorigo, M., & Gambardella, L. M. (1997b). Ant colony system: A cooperative learning approach to the traveling salesman problem. IEEE Transactions on Evolutionary Computation, 1(1), 5366.

Y.J. Yang, C. Wu / Expert Systems with Applications 36 (2009) 30343047 Stewart, C., Cristea, A., Brailsford, T., & Ashman, H. (2005). Authoring once, delivering many: Creating reusable adaptive courseware. In Webbased education conference WBE05, Grindelwald, Switzerland. Stu tzle, T. (1999). Local search algorithms for the combinatorial problems: Analysis, improvements, and new applications. DISKI (vol. 220). Sankt Augustin, Germany, Inx. Stu tzle, T., & Hoos, H. H. (1996). Improving the ant system: A detailed report on the MAX -MIN Ant System. Technical report AIDA-96-12, FG Intellektik, FB Information, TU Darmstadt, Germany. Stu tzle, T., & Hoos, H. H. (2000). MAXMIN Ant System. Future Generation Computer System, 16(8), 889914.

3047

Wang, T.-I., Wang, K.-T., & Huang, Y.-M. (2007). Using a style-based ant colony system for adaptive learning. Expert Systems with Applications. Available online. Weber, G. (1999). Adaptive learning systems in the world wide web. In: J. Kay (Ed.), Proceedings of the international conference on user modeling, UM99, Ban, Canada, June 2024, 1999 (pp. 371378). Wien, NewYork: Springer. Yang, Y. J., & Song, K. (2006). FOAPA: A fuzzy ontology e-education adaptive portal architecture for e-learner. In ICS 2006 proceedings of the international computer symposium 2006 (pp. 949957).