Amjad Mahmood
Southwest Jiao Tong University, China
Title: Semi-supervised evolutionary ensembles for web video categorization
Biography
Biography: Amjad Mahmood
Abstract
In recent decades, with the substantial growth in internet, computer hardware, infrastructure and optics have revolutionized the digital universe. On social web, multimedia advancement in digital world has provided an easy path to produce abundant videos by its users. This abundance of videos has made the selection criteria quite complicated for a user to search and get the desired video. Automatic Web Video Categorization (WVC) is principally a procedure of assigning web videos to pre-defined categories (such as Sports, Autos & Vehicle, Animals, Education, etc.). The challenges started from abundant data diversity within a category, deficiency in precisely labeled training data and atrociousness of video quality that make the analysis of diverse Web videos a challenging task. A lot of research work has been carried out on this issue in a conventional way by utilizing visual, textual and audio features individually or with different combinations to train models for Web video classification. Due to inadequate results of high-level concept detection methods and expensive feature extraction, the content based categorization could not achieve the required results. Basic text oriented mining approaches are quite ineffective to well categorize the YouTube videos due to inadequate textual details, e.g., title, tag and description. Besides supervised learning (classification), where some amount of pre-labeled data is used for learning purposes, unsupervised learning (clustering) has a unique role in data mining research. Evolutionary Algorithms (EA) have been developing rapidly as a powerful and general learning approach which has been used successful in finding a reasonable solution for data mining and knowledge discovery. Genetic algorithm (GA) is a kind of mainstream EA paradigm with a purpose of developing solutions for optimization problems. Clustering ensembles have emerged as an outstanding algorithm in machine learning to leverage the consensus across multiple clustering solutions and combines their predictions into a single solution with improved robustness, stability and accuracy. In this research work, we propose a Semi-Supervised Evolutionary Ensemble (SS-EE) frame work for social media mining, e.g., Web Video Categorization (WVC), using their low cost textual features, intrinsic relations and extrinsic Web support.