Frequent temporal pattern mining incrementally from educational databases in an academic credit system

Abstract

Educational data mining is emerging for useful knowledge hidden in educational databases. Frequent temporal pattern mining is one of the popular mining tasks to help us get insights into the characteristics of the students and further of their study. As time goes, educational databases in an academic credit system keep increasing and updated in nature. Thus, frequent temporal pattern mining in educational databases needs to carry out in such a non-trivial situation. At present, it is found that many sequential mining techniques just considered a sequence of ordered events with no explicit time and most of the temporal pattern mining techniques handled temporal databases where temporal information is associated with each transaction, whereas those discovering frequent temporal patterns with timestamped elements were not proposed for incremental databases. In order to achieve frequent temporal patterns in educational databases that contain timestamp-extended sequences, our work defines a temporal comprehensive incremental sequential pattern mining algorithm, TCISpan, based on prefix trees to organize frequent temporal patterns for efficiently mining incremental educational databases along the time. Experimental results on both real and synthetic datasets have shown that the proposed algorithm outperforms the mining approach that conducts the mining task on incremental timestamp-extended sequence databases from scratch.