Clustering techniques for web mining

Abstract

The project mainly focuses on the design and implementation of two newly proposed data clustering algorithms, namely Active Fuzzy Constrained Clustering (AFCC) and Semi-Supervised Heuristic Fuzzy Co-clustering with the Ruspini’s condition (SS-HFCR), for semi-supervised data analysis.

The Active Fuzzy Constrained Clustering (AFCC) technique tries to take into account simple yet useful constraints provided by the user to “steer” the clustering process to produce more preferred results. Not only AFCC, seeing the possibility of enhancing the newly proposed algorithm namely Heuristic Fuzzy Co-clustering with the Ruspini’s condition (HFCR) with the same fashion as AFCC, an improved Semi-Supervised HFCR (SS-HFCR) is then developed.

After being successfully implemented both algorithms in Java, experimental studies have been conducted to both verify the coding and the performance of both algorithms. As to qualify the performance, analyses on the precision, recall and purity measurement were carried out.