A Shortest Learning Path Selection Algorithm in E-learning Chengling Zhao and Liyong Wan Department of Information Technology, Central China Normal University, Wu-Han, China [email protected] Abstract Generally speaking, in the e-learning systems, a course is modeled as a graph, where each node represents a knowledge node (KU) and two nodes are connected to form a semantic network. The desired knowledge is provided by the student as a direct request or from search results, mapping the owned knowledge onto the target knowledge. How to select a learning path which costs the least time and effort is critical. In this paper, we described the relationships between different nodes in the graph structure of knowledge units and propose an algorithm to select the shortest learning paths to learn the target knowledge. 1. Introduction With the development of the Internet, e-learning system has become more and more popular because it can make learner study at any time and any location conveniently. In the e-learning systems, a course is modeled as a graph, where each node represents a knowledge unit (KU) and two nodes are connected only if the knowledge gained by learning the former node is needed as a prerequisite to learn the latter node [1]. A KU includes the course theory, principle, concept, definition, worked example and exercise, etc. the relationship between two KUs can be divided into three types: precedence relationship, succession relationship and parallel relationship. The desired knowledge is provided by the student as a direct request or from search results, mapping the owned knowledge onto the target knowledge. In this case the student may have a number of learning paths to reach the target knowledge. How to select a learning path which costs the least time and effort is critical. 2. Learning path and arrowhead weight 2.1. Learning path All KUs belonging to the same course are connected together into a graph structure using oriented arrowheads; the orientation represents the three types of relationships between two units. The target is to create and propose to the student all the possible paths starting from his/her knowledge and directed to the desired topic of interest according to the connections between different nodes. A graph structure of KUs is illustrated in figure 1. Figure 1. A graph structure of KUs. From figure 1 we can see that there are 10 KUs in the graph. 6 and 9 form a precedence relationship; 1 and 2 form a succession relationship; 7 and 8 form a parallel relationship. We suppose the target KU is the node 6 and the student owns the knowledge node 2. Figure 2 shows all paths from the node 2 to the node 6. Figure 2. All the paths from the node 2 to the node 6. 2.2. Arrowhead weight When we list all learning paths from the KU the learner owns to the KU the learner want to reach, we must consider an issue: which learning path is the best for the learners. We define the best learning path is the learning process through which will cost the least time and effort. Then we should consider weight of an arrowhead represents the difficulty to access a topic coming from a previous one. Arrowhead weights are managed by teachers, usually by establishing their initial values, which are further refined with the system help, e.g. through a statistical analysis of previous students sessions [1]. Each oriented arrowhead Proceedings of the Sixth International Conference on Advanced Learning Technologies (ICALT'06) 0-7695-2632-2/06 $20.00 © 2006 IEEE connecting two nodes is weighted in order to measure the difficulty to face the next node. Here we suppose the average weight between two nodes is 1.0. The higher value of weight means that the learner needs more time and effort to learn the next KU. Figure 3 shows an example of graph consists of arrowhead weights. Figure 3. An example of graph consists of arrowhead weights. The weights are beside the arrowheads. If the student have learned node A, but B, C, D and E have not yet learned. Now his target KU is E, obviously there have been 9 learning paths: A-B-E, A-B-C-E, A-B-C-D-E, A-C-E, A-C-B-E, A-C-D-E, A-D-C-E and A-D-C-B-E. If we select A-B-E, weights= ¦ weight =0.7+1.1=1.8; if we select a) If vij=i , the shortest path is vi to vj; b) If vij z i , we suppose vij=k, the shortest path is vi –vk- vj ,then if vik=i, it means that there is no medial node between vi and vk; if vkj=k , it means that there is no medial node between vk and vj , else we can find a medial node k. c) Repeat a) and b) until we find out all of the medial nodes. d) Connect vi and vj with these medial nodes, we can get the concrete shortest learning paths (one path or more than one). Now we use the algorithm to select the shortest learning path in figure 3. we suppose the IDs of A,B,C, and D are:1,2,3,4 and 5. So D= M 0.7 1.2 1.0 M V= § 1 1 1 1 1 · § ¨ ¨M ¨M ¨ ¨M ¨M © · ¸ M 0.3 M 1.1 ¸ 0.4 M 0.2 0.9 ¸ ¸ M 0.6 M 0.8 ¸ M M M M ¸¹ 3. The algorithm We suppose a graph structure consists of n nodes, and then the algorithm is below: (1) Construct a initial adjacency matrix D=pij , pij is the arrowhead weight from node i to node j. if there is no arrowhead exist between the two nodes or the orientation of the arrowhead is reverse or i=j, pij=M. [2] (2) Construct a medial node matrix V=vij , the value of vij is the ID of node i. (3) Start iterative operation˄the times of iterative operation equal to the number of nodes. we first compare pij (i, j=2,3…, n)with pi1+p1j, there are three cases: a) If pij> pi1+p1j, replace pi1+p1jwith pij in the next adjacency matrix, and vij=1; b) If pij<pi1+p1j, there is no change in the next adjacency matrix and node matrix; c) If pij=pi1+p1j, there are two results, one is as a) and the other is as b) So we get the next adjacency matrix D1 and node matrix V1, then begin the next iterative operation until we get Dn and Vn . (4) Search for the shortest path from vi to vj, ¨ ¨M ¨M ¨ ¨M ¨M © IEEE ¸ 0.7 0.3 0.5 1.1 ¸ 0.4 0.7 0.2 0.9 ¸ ¸ 1.0 0.6 0.8 0.8 ¸ M M M M ¸¹ The other case is D5= §M ¨ ¨M ¨M ¨ ¨M ¨M © 0.7 1.0 1.0 1.8 · ¸ 0.7 0.3 0.5 1.1 ¸ 0.4 0.7 0.2 0.9 ¸ ¸ 1.0 0.6 0.8 0.8 ¸ M M M M ¸¹ ¨2 ¨3 ¨ ¨4 ¨5 © 3 2 3 2¸ 3 2 3 3¸ ¸ 3 4 3 4¸ 5 5 5 5 ¸¹ V 5= § 1 ¨ ¨2 ¨3 ¨ ¨4 ¨5 © 1 3 3 3 2 2 2 4 1 3 3 3 2· ¸ 2¸ 3¸ ¸ 4¸ 5 5 5 5 ¸¹ We can see that p15=1.8 (A-E), then we search for the medial nodes can get the shortest path is 1-2-5 (A-B-E) and 1-4-5(A-D-E). 4. Conclusions and future work In e-learning courses, how to select a learning path which costs the least time and effort is critical. In this paper, we describe the relationships between different nodes in the graph structure of knowledge units and propose an algorithm to select the shortest learning paths to learn the target knowledge. In the future, we will use this algorithm to construct an adaptive learning system, which can provide the best and easiest learning experience to the learners. 5. References [1]Vincenza Carchiolo, Alessandro Longheu, Michele Malgeri. Adaptive Formative Paths in a Web-based Learning Environment. Educational Technology & Society. 4(2002), pp.64-75. [2] Wang Ying. Algorithm Research of the Optimal Path in Intelligent Traffic System. Computer Development & Applications.5(2005), pp. 45-47. Proceedings of the Sixth International Conference on Advanced Learning Technologies (ICALT'06) 0-7695-2632-2/06 $20.00 © 2006 2 2 2 2 ¸¸ 3 3 3 3¸ ¸ 4 4 4 4¸ 5 5 5 5 ¸¹ Through 5 times of iterative operation, the adjacency matrix and medial node matrix are as follows: One case is D5= § M 0.7 1.0 1.0 1.8 · V5= §¨ 1 1 2 1 4 ·¸ A-C-B-E, weights= ¦ weight =1.2+0.4+1.1=2.7. The smallest weights mean the shortest path. In the next section we will propose an algorithm to select the shortest learning path. ¨2 ¨ ¨3 ¨ ¨4 ¨5 ©
© Copyright 2026 Paperzz