A Weighted Distance Metric Clustering Method to Cluster Small Data Points from a Projected Database Generated from a Freespan Algorithm

Background/Objectives: Clustering and Sequential Pattern Mining is two most important unsupervised learning algorithms. The objective is to mine small projected databases rejected by Frequent Pattern - Projected Sequential Pattern mining (FreeSpan) technique using a weighted distance metric clustering method, a process of finding the distance between the small data points and cluster it so that it cannot be rejected. Methods/Statistical Analysis: The method involves the implementation of a distance metric clustering algorithm over a FreeSpan technique to cluster the data points of small projected databases. The FreeSpan technique can be considered as an ensemble of clustering and sequential pattern mining methods. Findings: The clustering method clusters the data points resulted from the FreeSpan technique that are ignored after the scanning process as their sizes are very small. The clustered data therefore gathers the ignored data points thereby providing an accurate clustered data containing small data points which results is trustable sequential pattern for future predictions. The proposed system reduces the complexity by incorporating just a single clustering algorithm. Therefore the major operations of the algorithm remain undisturbed and give its efficient output and also the output is found to be accurate and stable. Applications/Improvements: The technique proposed in the paper can be applied to datasets that needs to be clustered for decision making. The same technique holds good and can be made applicable to high dimensional views.