Page 167 - DCAP606_BUSINESS_INTELLIGENCE

Business Intelligence

Notes
ITS with Proactive Feedback
Data mining findings can also be used to improve the tutoring system. We implemented
a function in Tada-Ed allowing the teacher to extract patterns with a view to integrate
them in the ITS from which the data was recorded. Presently this functionality is available
for Association Rule module. That is, the teacher can extract any association rule. Rules are
then saved in an XML file and fed into the pedagogical module of the ITS. Along with the
pattern, the teacher can specify an URL that will be added to the feedback window and
where the teacher can design his/her own proactive feedback for that particular sequence
of mistakes C (which the student has not yet made).

(a) XML encoded patterns (b) Screen shot of mistake viewer
The structure of the XML file is fairly simple and is shown in (a). For instance, using our
logic data, we extracted the rule saying that if a student makes the mistakes “Invalid
justification” followed by “Premise set incorrect” then she/he is likely to make the mistake
“Wrong number of references lines given” in a later step (presently there is no restriction
on the time window). This rule has a support of 47% and a confidence of 74%. The teacher,
when saving the pattern, also entered an URL to be prompted to the student. The pedagogical
module of the Logic Tutor then reads the file and adds the rule to its knowledge base.
Then, when the student makes these two initial mistakes, she/he will receive, in addition
to the relevant feedback on that mistake, an additional message in the same window (in a
different colour) advising him/her to consult the web page created by the teacher for this
particular sequence of mistakes.
Support for Student Reflection
Extracting information from a group of learners is also extremely relevant to the learner
themselves. The fact that learner reflection promotes learning is widely acknowledged.
The issue is how to support it well. A very useful way to reflect on one’s learning is to look
up what has been learned and what has not yet been learned according to a set of learning
goals, as well as the difficulties currently encountered. We are seeking here to help learners
to compare their achievements and problems in regards to some important patterns found
in the class data. For instance, using a decision tree to predict marks, the student can
predict his/her performance according to his/her achievements so far and have the time
to rectify if needed. Here more work needs to be done to assess how useful this prediction
is for the student.
Questions:
1. Discuss how the discovery of different patterns through different data mining
algorithms and visualisation techniques suggest you a simple pedagogical policy?
2. Also discuss the behaviour of clustering and cluster visualisation.
Source: books.google.co.in/books?isbn=1586035304

162 LOVELY PROFESSIONAL UNIVERSITY