Clustering techniques for web documents

• Main Author: Pan, Tianchi
• Other Authors: Chen Lihui
• Format: Final Year Project
• Language: English
• Published: 2013
• Subjects: DRNTU::Engineering
• Online Access: http://hdl.handle.net/10356/54272
• Institution: Nanyang Technological University

Document clustering is a process of grouping documents into several natural and homogeneous clusters so that documents within the same cluster are more similar to each other than those belonging to other clusters [1]. While in the web environment, task seems more challenging. Essential clustering techniques need to be employed to facilitate the discovery knowledge in this process. K-means is one of the frequently used methods in data clustering; however, it will fail to find out the meaningful clustering result if input data is given in a less structured way. Therefore, in this report a new learning distance metric proposed by Eric P. Xing is implemented with supplementary side information to help improving the K-means clustering performance. New algorithm will be studied in details and validated on different datasets and its performance will be evaluated by some quantitative values: NMI, purity and random index using Java as well as cluster visualization using MATLAB. From the results obtained, we have found that new clustering algorithm has shown a pleasant improvement compared with the original one and might be used for future application in data clustering.