Jefferson Lai and Benjamin Le and Pierce Vollucci and Wenxuan Cai and Yaohui Ye

EECS Department, University of California, Berkeley

Technical Report No. UCB/EECS-2015-72

May 13, 2015

http://www2.eecs.berkeley.edu/Pubs/TechRpts/2015/EECS-2015-72.pdf

This capstone project report covers the research and development of Smart Anomaly Detection and Subscriber Analysis in the domain of Online Video Data Analytics. In the co-written portions of this document, we discuss the projected commercialization success of our products by analyzing worldwide trends in online video, presenting a competitive business strategy, and describing several approaches towards the management of our intellectual property. In the individually written portion of this document, we discuss our implementation of two machine learning models, k-Nearest Neighbors and Random Forest, and evaluate them as a means of identifying subscription churners, the primary goal of Subscriber Analysis. In particular, we show that the performances achieved by these models using our initial, restricted feature set are promising and warrant future exploration of these models.

Advisors: George Necula

---

BibTeX citation:

@mastersthesis{Lai:EECS-2015-72,
    Author= {Lai, Jefferson and Le, Benjamin and Vollucci, Pierce and Cai, Wenxuan and Ye, Yaohui},
    Editor= {Necula, George and Wroblewski, Don},
    Title= {Online Video Data Analytics},
    School= {EECS Department, University of California, Berkeley},
    Year= {2015},
    Month= {May},
    Url= {http://www2.eecs.berkeley.edu/Pubs/TechRpts/2015/EECS-2015-72.html},
    Number= {UCB/EECS-2015-72},
    Abstract= {This capstone project report covers the research and development of Smart Anomaly Detection and Subscriber Analysis in the domain of Online Video Data Analytics. In the co-written portions of this document, we discuss the projected commercialization success of our products by analyzing worldwide trends in online video, presenting a competitive business strategy, and describing several approaches towards the management of our intellectual property. In the individually written portion of this document, we discuss our implementation of two machine learning models, k-Nearest Neighbors and Random Forest, and evaluate them as a means of identifying subscription churners, the primary goal of Subscriber Analysis. In particular, we show that the performances achieved by these models using our initial, restricted feature set are promising and warrant future exploration of these models.},
}

EndNote citation:

%0 Thesis
%A Lai, Jefferson 
%A Le, Benjamin 
%A Vollucci, Pierce 
%A Cai, Wenxuan 
%A Ye, Yaohui 
%E Necula, George 
%E Wroblewski, Don 
%T Online Video Data Analytics
%I EECS Department, University of California, Berkeley
%D 2015
%8 May 13
%@ UCB/EECS-2015-72
%U http://www2.eecs.berkeley.edu/Pubs/TechRpts/2015/EECS-2015-72.html
%F Lai:EECS-2015-72