INCREMENTAL MINING OF FREQUENT PATTERNS FROM THE EDUCATIONAL DATA

Abstract

Changes in the underlying database occur due to proliferation of innovations from heterogeneous online and offline resources. Mining frequent patterns are costly in changing databases, since it requires scanning the database from the start. Incremental Mining of frequent patterns is a solution to deal with this problem. The incremental mining process uses previous mining result to get the desired knowledge by reducing mining costs in terms of time and space. This research focuses on incremental mining of frequent patterns using four approaches. Firstly, it is important to take into account the functionality i.e. Kind of patterns and performance (how to mine frequent patterns) of the patterns to be mined. One approach to find frequent patterns from frequently changing databases is to construct efficient data structures. A novel tree based data structure and mining algorithm called TIMFP (Tree for Incremental Mining of Frequent Pattern) is proposed, which is compact as well as based on “Build Once and Mine Many” property. Secondly, to measure the semantic significance of an item as per users‟ perspectives, a novel approach for mining high utility is presented. Focus of this research work is to introduce Average Maximum Utility (AvgMU) concept, which prunes items in early stages to avoid unnecessary staying of the items in the pruning phases. Thirdly, this work proposes a tree based algorithm called CIFMine (Constraint based Incremental Frequent Pattern Mining) to mine the incremental data by using a dataset filtering technique. In general, the goal of the pattern mining process is to

v discover all the patterns from the data sources where not all the patterns may be suitable for the end user. This approach focuses on the constraints and specifies the desired properties of the patterns to mine that are likely to be of interest for the end user. Lastly, we have focused on Educational Data Mining (EDM) by focusing research trends, challenges and Educational Outcomes to enhance the quality of education. A modified constraint based algorithm “Modified TIMFP” (Modified Tree for Incremental Mining of Frequent Pattern) to construct and mine incremental educational data is proposed. All the approaches are tested using Synthetic and Real dataset. This research successfully establishes the fact that, approaches for Incremental Mining of frequent patterns are cost effective (in terms of time and space), efficient, scalable and produces optimal solutions for mining frequent patterns from different kind of data as compared to traditional frequent pattern mining approaches.