Big data analysis on online TV – Predict likes in FaceBook

Author Name(s): Arunkumar K, Dr. Devendran A
Author Email: arunkumar@gmail.com

Abstract

Online video is a powerful way to communicate or express any brand story, explain business value proposition clearly, and build relationships with your customers and prospects better than any other means. Recent statistics shows that including video on a landing page can increase conversion rates by 80%? Or that 92% of mobile video consumers share videos with others. Other stats show video is shared 1200% more than both links and text combined. Hence, video content is a powerful tool for any brand. The most popular ways to take video online are YouTube channels and Facebook pages. It is critical for business to track how their videos are performing. Both Facebook and YouTube provide enormous data on videos in their platform from video views, unique video views, the average duration of the video view and audience retention. We studied related work done in predicting no. of likes in Facebook and experimented our new predicting system on various scenarios and observed lot of interesting facts.

Keywords

Bigdata , Online Tv, Facebook , Videos

Introduction

Past decade shows an increase in the no. of video consumed over internet and hence an increase in no.of online channels. According to Cisco Forecast, video will represent 69% of all consumerbasedInternet traffic by 2017; this is expected to rise to 80% by 2019. We have YouTube and Facebook having major video delivery share. Also, we see some new online video platforms such as Vine, Instagram, and Snapchat. To give a feel of how big these video numbers are, here are some of official numbers from YouTubeand Facebook. In January 2016, Facebookannounced that people watch around 10 million hours ofvideo a day. In February 2016, Google CEO, SundarPichai, reminded investors that YouTube’saudience watches hundreds of millions of hours of video every day.

Online TV: Online television (or internet television) is the digital distribution of television content, such as TV shows, via the public Internet (which also caries other types of data), as opposed to dedicated terrestrial television via an over-the-air aerial system, cable television, and/or satellite television systems. The amount of
data these channels put online is really huge and their success mainly depends on how these data is consumed by their subscribers. Both Facebook and you tube gives lot of data related to the views. It is crucial to analyze these data to have business insights.

Related Works The closest work we could find in similar area is done by Kevin at al [1] in predicting the popularity of a Facebook status based on that user’s personal Facebook data. Popularity, measured by the number of “likes” a post gets, is important to users because it eases the anxiety that users feel when sharing on social networking sites. Facebook is one of the core ways people express themselves, but users often struggle to decide whether their thoughts are worth sharing.

Conclusion

From the above observation, following is the list of varying factors and their influences. At this point we concluded that actual content plays a major role in popularity along with day and timeof post. Hence, we build our Regression model around these 3 parameters with more weighting on content.

We have built our recommendation system predicting the no. of likes, shares and reach for newcontents to be posted. This gave us god accuracy for selected online channel. We also observed this is not same for all the channels. Hence, we built an analysis tool and allowed the user (channel owners) to view and chose parameters for the recommendation system and view prediction accuracy.

141 total views, 1 views today

Download PDF File

About the author: admin