Don't forget to include the attribute generally considered by lecturers to be the most effective predictor in your dataset:

The likelyhood of a student progressing / passing is inversely proportion to the distance (given a free choice) the student chooses to sit from the lecturer in the lecture theatre :)

Educational data mining methods belong to a diversity of literature including data mining, machine learning, information visualization, and computational modeling. Machine learning approaches include neural network, naive Bayesian, K-nearest neighborhood, decision tree, support vector machine (SVM), linear regression, and rule induction.

http://waset.org/publications/16952/an-educational-data-mining-system-for-advising-higher-education-students

I've seen decision trees to outperform naive Bayes but not specifically on SAP data.

I think the ANN can solve this problem. You can try.

Thank you Nikola, yes decision trees have become more favourable over the years, but considering that educational data does not have the same distinct benchmarks that abound in business related data, I am yet to see or hear of where decision trees have been implemented on educational data especially for SAP. It is however a strong contender, that much I agree.

Prof Krishnan thanks for your expose and literature sent. I do however believe that some techniques such as the linear regression may not be suitable for SAP.

Thank you Yancang, ANN may suffice and is worth a try as well.

ANN will be most effectively use on SAP. try that

Don't forget to include the attribute generally considered by lecturers to be the most effective predictor in your dataset:

The likelyhood of a student progressing / passing is inversely proportion to the distance (given a free choice) the student chooses to sit from the lecturer in the lecture theatre :)

Any empirical evidence to prove your theory James? :)) 

Actually it's the aim of my area research, now we implemented  a framework named LearnFitII using collaborative filtering as a datamining tool to help learner and to enhance learning performance this result of this work will be published soon

Apart from the oral tradition handed down by generations of previous lecturers, I did a study of 3 years attendance data from computing and software engineering students on HND programmes. The best indicator was that when sorted by order of attendance, virtually all of the students who passed were above the 80% attendance boderline and virtually all of the students who failed were below it (with some noise around the boundaries). Students who do not attend are, of course, choosing to sit further from the lecturer thasn those who do! Second best predictor  a colleague found, which could be applied before admission to courses, was that students who chose Science and Engineering subjects at school progressed better than students who studied Arts & Humanities subjects, regardless of UCAS point scores. Also an very good predictor of failure was if a student doing science subjects had originally applied for a pharmacy course. All this points up the value of the Elementary  Data Analysis phase of designing an AI classifier, which supports good feature extraction - so there is a bit of a serious point to my light-hearted comment.

James, can you kindly furnish me a copy of this research? 

Essaid, I eagerly await your published work as this also is part of my PhD work

Dantala: In your review of existing literature, did you at all gather why the Naive Bayes Classifier (NBC) has been more extensively exploited for the SAP prediction task? If it has been more widely explored, perhaps there are underlying arguments why it is the more preferred method in comparison to others. It would help you to explore this aspect fully before embarking on exploring heretofore unexplored or little explored techniques. 

For it seems that BNC is a technique easy to impliment for potentiel SAP  but we have to take in mind that the profil can be well modelled by adding also the psychological technique like MBTI Model or Kolb Model to deal eventulally with somme datamining problems. Actulally I will send a draft and look this work

http://scholar.google.com/citations?user=ZAxOomkAAAAJ&hl=en

Dantala, I would always try an absolutely straightforward backprop ANN (as we find a well trained backprop ANN is a good general technique which provides a baseline for measuring improvement with other classification or regression techniques). Also if you are going to make decisions about students based on this technique you need some kind of explanation of how the decision was reached in order to justify it (there's a good paper on the need for explainability in black box AI systems controlling student option choices lurking out there somewhere, but I can't lay my hands on it at the moment - you may be able to find it with either Google Scholar or Science Direct, but you are searching for transparency of AI classifiers rather than data mining to find it). Therefore I suggest using a decision tree. We've looked at domains where Bayesian techniques are popular in the past and sometimes it's because the investigators are statisticians rather than AI practitioners at heart, and so are more comfortable with Bayesian. I conducted some experiments on classifying Dialogue Acts where Bayesian techniques dominate and found decision trees to be more effective with my problem.

You may want to take a look at the works of Thai-Nghe (https://sites.google.com/site/ntnghe/)

http://dblp1.uni-trier.de/pers/hd/t/Thai=Nghe:Nguyen

e.g. http://ceur-ws.org/Vol-872/factmod2012_paper_2.pdf

And his PhD thesis entitled "Predicting Student Performance in an Intelligent Tutoring System":

http://d-nb.info/102380946X/34

n my research area related to how we can enhance learning by using datamining algorithms. In order to compare our algorithms we need a real dataTel and unfortunatly we could'nt find it.So what we find is just dataset related to ebusiness (BookCrossing, Movies,...). Can some one help us if there is any dataTel available in the net ?Ask

Fuzzy Set Theory, Rough Set Theory

@ Essaid

Similar ot DataTel data base creation, there is an information system for students (free).

Student Information Management System Software

http://megaupl0ad.ws/files/student%20information%20management%20system%20software%20free&id=my