Web Video Object Mining: Expectation Maximization and Density Based Clustering of Web Video Metadata Objects

Nowadays YouTube becoming most popular video sharing website, and is established in 2005. The YouTube official website is providing different categories videos including Science and Technology, Films and Animation, News and politics, Movies, Comedy, Sports, Music etc. Each video hosted in website such as YouTube have its own identity and features. The identity and features of each video can be described by web video metadata objects such asURL of each video, category, length of the video, rating information, view counts, comment information, key words etc. Using extracted web video metadata objects, we present an indepth and systematic clustering study on the metadata objects of YouTube videos using Expectation Maximization (EM) and Density Based (DB) clustering approach. Distinct web video metadata object clusters are formed based on different category of web videos. The resultant clusters are analyzed in depth as a step in the KDD process.