Knowledge Discovery from E-Learning Activities Table 12 contains the six first sorted contributions of Web activities of the ICA mixture matrix for the 5 sources estimated. Each source was associated with one learning dimension of Table 1 analyzing the weight of the Web activities and considering the principal evaluation methodologies employed by teachers for graduate courses with grades. Dimension 1 was not detected and dimension 5 was detected twice. The methodologies assigned grades focusing on: achievement, individual student participation, or group work. The implicit teaching styles of the evaluation methodologies encourage specific learning styles of the students, as we explain below. The learning dimension 1 (sensory-intuitive) corresponding to “perception” was not detected in the ICA mixing matrix; it could be because the emphasis of educational strategies did not favour to highlight that dimension. From Table 12, the relationship between learning style dimensions and Web activities can be made; see Table 13 where we have added a possible Web activity combination for learning dimension 1. Note that some Web activities are associated with more than one dimension; it has sense because a Web activity could demand several capabilities of the students used in their learning process. Allowing that kind of relationship we can obtain more real and versatile descriptions of the student learning styles, besides of including all the dimensions of the learning framework. In Garcia et al., (2007) just three dimensions of the Felder and Silverman (1988) model were considered and the Bayesian network proposed constrained Table 12. ICA mixing matrix ( * learning style dimension, ** workgroup documents) LSD * 2 4 3 5’ 5 Sorted Web activity contribution chat forum news e-mail access exercises 1 .82283 .30755 .16476 .14756 .14231 e-mail content wg-doc ** exercises forum chat 1 .34189 .32297 .28768 .22548 .20078 wg-doc news achieve content chat e-mail 1 .80531 .4122 .39987 .39421 .31666 achieve content agenda access forum news 1 .45124 .2117 .21116 .20087 .18239 access agenda content achieve e-mail chat 1 .95776 .85549 .7143 .5832 .49774 Table 13. Association between learning styles and Web activities Learning Style Web event activity 1 Sensory-Intuitive Perception chats, forum participation, course access. 2 Visual-Auditory Input 3 Inductive-Deductive Organization 4 Active-Reflective Processing 5 Sequential-Global Understanding chats, forum participation, news reading, e-mail exchange. workgroup document, news reading, course achievement, content consulting . e-mail exchange, content consulting, workgroup document, exercise practice. course access, agenda using, content consulting, course achievement.
Knowledge Discovery from E-Learning Activities relationship of the Web activities with just one dimension of the learning model. Figure 6 shows the sources 3, 4, and 5 (organization, processing, understanding) obtained for the grade graduate course dataset. Four labelled characterised zones in the learning style space are displayed: (1) Represents the learning style more important in the population. The learning for the students in this zone emphasizes global understanding, active processing, and deductive logic (natural human teaching style), and high grades. (2) This learning style is focused on inductive logic (natural human learning style), with sequential understanding, and relative active processing. Students within this style could have natural skills for virtual education. (3) It is characterised by global understanding, deductive logic, and reflective processing. Students within this style would have higher abstraction skills that need of teaching. (4) Basically this cluster represents outliers with individual learning styles. We can conclude that dimension of understanding enables to project clearly the learning styles, and its principal components are achievement, content, and agenda. This finding confirms the assumption that the more quickly way to change the learning style of the student is to change the assessment style, that is, expected evaluation bias how the student learns (Elton & Laurillard, 1979). We made a cluster validation procedure to determine best quality of cluster configuration for data of Figure 6. It consisted in estimating the partition coefficient and the partition entropy coefficient for different number of clusters (Haldiki, Batistakis, & Vazirgiannis, 2001). The best cluster configuration for data of Figure 6 was 4 clusters—a detailed explanation of cluster validation procedure is in cluster analysis section. Figure 7 shows three sources for graduate courses with no grades. The distribution of the data in Figure 7 does not allow forming learning style groups and show all the subjects within a unique learning style. As understanding and organization dimensions do not discriminate projection of the learning styles, only the dimension of the processing provides some discrimination. The unique learning style emphasises reflection over actuations. It would be the content consulting and exercise practice components of that dimension. The conclusion is the lack of assessment does not allow developing student learning styles. Results for regular academic career courses were similar to the graduate courses results finding meaningful learning styles for courses with grades. Results of this section could be analyzed as a kind of ontology. The conceptions are the learning styles detected, related with the dimensions Figure 6. Three sources in a learning style space for graduate courses with grades