Page 1 of 7
European Journal of Business &
Social Sciences
Available at https://ejbss.org/
ISSN: 2235-767X
Volume 07 Issue 04
April 2019
Available online: https://ejbss.org/ P a g e | 64
Predicting the Future of Streaming Videos
with Popularity and Factors
M Debora, M Chalapathi Rao
B. Tech Student, Department of CSE, CMR Technical Campus,Kandlakoya Village, Medchal
District, Telangana, India.
Associate Professor, Department of CSE, CMR Technical Campus,Kandlakoya
Village, Medchal District, Telangana, India.
Abstract - A huge User Generated Content is
re-shaping the way people watch video and TV,
with millions of video producers and consumers.
In particular, User Generated Content sites are
creating new viewing patterns and social
interactions, empowering users to be more
creative, and developing new business
opportunities. To better understand the impact of
User Generated Content systems, we will
analyse the video streaming websites. Based on
a large amount of data collected, we provide an
in-depth study of User Generated Content
systems.
In this paper we will design a architecture which
will analyse the User Generated Content and
will Predict the future of the streaming Videos
based on the characteristics of the user data
(like type of Video, content of video, time of the
video, genre of the video etc..). This helps the
video provides predict their future and helps in
improving the content based video as per the
users. Overall, we believe that the results
presented in this paper are crucial in
understanding User Generated Content systems
and can provide valuable information to ISPs,
site administrators, and content owners with
major commercial and technical implications.
Keywords: Data analysis, data mining,
measurement, prediction..
I INTRODUCTION
With the ubiquitous access of Internet and the
emergence of Web 2.0 services, an enormous
and ever growing amount of online content has
been brought into the digital world. Content
producers now can reach audiences in
inconceivable numbers that are unmatched
through conventional channels. Among the
various kinds of online content, online videos
are currently a dominating component of the
Internet. In terms of bytes, the video traffic
accounts for around 64% of all the Internet
traffic in 2014, and will be up to 80% by 2019
[1]. This explosive growth intensifies the online
competition for user attention, and contributes to
a ‘‘winner-take-all’’ online video ecosystem: a
Page 2 of 7
European Journal of Business &
Social Sciences
Available at https://ejbss.org/
ISSN: 2235-767X
Volume 07 Issue 04
April 2019
Available online: https://ejbss.org/ P a g e | 65
small fraction of videos attract most of the user
interest, whereas the vast majority of videos are
of limited views [2], [3]. Given the huge amount
of video content and the high variability of user
attention, it is of utmost importance for a number
of tasks to understand the characteristics of
online video popularity and further predict the
popularity of individual videos. For service
providers, the video popularity dynamics and
prediction results can greatly benefit their future
design of the content filtering, video ranking,
and recommendation schemes, which help users
to find videos with more potential values more
easily[4] .For advertisers in the online
marketing, prediction of the next rising star of
the
Internetcanmaximizetheirrevenuesthroughbettera
dvertisingplacement[3].With the extrapolation of
video popularity, network operators can
proactively manage the bandwidth
requirementanddeploythecacheserversintheconte
ntdeliverynetwork (CDN) for hot videos in
advance[5]. In addition, in a quite different
context, the video popularity will be of great
interest in the opportunistic communications
among mobile devices [6]. In such
resource-constrained environments (e.g. limited
bandwidth and storage space), predicting hot
videos is helpful for the content delivering,
caching and replicating on the device end. In this
paper, we study the video popularity of Youku
(www.youku.com), a leading online video
service provider in China. Our work is based on
the data of 33,359 videos crawled from Youku
website for 30 consecutive days. With these
data, we analyze in-depth how the popularity of
online video content evolves overtime, and how
to predict the future popularity of an individual
video. The main contributions of our work ares
ummarized as follows:
1) We provide a detailed characterization of the
popularity dynamics of online videos. In
particular,we provide insights into the popularity
evolution patterns of the individual videos.
2) We tackle the problem of popularity
prediction by proposing a model that can capture
the popularity evolution of an individual
video.To the best of our knowledge, the
proposed method is the first to specialize models
by popularity evolution patterns in the popularity
prediction.
3) We evaluate our model on a real-world
dataset and compare the prediction performance
with two state of-the-art online video popularity
prediction models. Our approach leads to
significant reductions in prediction errors
of32.73%and11.28% over the base line models
respectively.
4) We analyze the potentials and limitations of
different detectors and model parameters in the
prediction. We shed light on the importance of
each feature and each feature group used in the
burst prediction.
.II RELATED WORK
A. ANALYSIS OF GROUP POPULARITY The
beginning research of video popularity can be
found in the early studies on user access
Page 3 of 7
European Journal of Business &
Social Sciences
Available at https://ejbss.org/
ISSN: 2235-767X
Volume 07 Issue 04
April 2019
Available online: https://ejbss.org/ P a g e | 66
patterns. Gill et al. [7] analyzed the user access
patterns, file properties, video view count
distribution and referencing behaviors of
YouTube, based on the traffic collected at a
campus. Zink et al. [8] also captured traffic from
a university campus and analyzed the duration,
data rate, population and access patterns of
YouTube videos. Both of them discovered that
the video requests of YouTube followed a
Zipf-like distribution. Chaetal. [4]analyzed the
popularity distribution, popularity evolution and
content characteristics of YouTube and a
popular Korean video sharing service. They
modeled the group video popularity as a power
law with an exponential cut-off.
Theyalsoinvestigateddifferentmechanisms,sucha
scaching and P2P, to improve the video
distribution. Cheng et al. [9] performed a
long-term crawling of YouTube and studied the
static properties, access pattern, popularity
distribution,
popularitytrendandsocialnetworkingofYouTubev
ideos.These previous studies provide important
insights into the traffic and content popularity of
online video service. However, they focused
mostly on the overall popularity distribution of a
group of videos, and they collected data at either
a single or several static time points (i.e.
snapshots). Our study complements these
existing works by further characterizing the
video popularity at a more fine-grained
individual level besides the group level.
Moreover, we track the popularity of each video
every day since their publications, and analyze
the popularity growth over time during the
whole
observation period. We also further tackle the
challenge of future popularity prediction in our
paper.
B. CHARACTERIZATION OF POPULARITY
EVOLUTION Some studies perform temporal
analysis of how video popularity evolves over
time, especially on the peak days, and
clustervideoswhichhavesimilarpopularityevoluti
ontrends. Crane and Sornette [10] first proposed
epidemic models to describe the popularity
evolution of YouTube videos. They
distinguished four different evolution patterns
based on the fraction of views on the peak day
and explained the sepatterns in terms of
endogenous and exogenous effects. Yang et al.
[11] studied how the popularity of online content
grew and faded over time, and proposed a
clustering algorithm based on the k-means
method to identify the temporal patterns of video
popularity. Figueiredo et al. [12]–[14]
characterized the popularity evolution patterns of
YouTube videos based on the classification
method in[10],and studied the impacts of
different types of referrers on such patterns.
Ahmedetal.[15] identified the patterns of
temporal evolution for distinct types of data
overtime and predicted the evolution pattern of
popularity in user generated content. Those
patterns proposed by the previous works can
well describe the evolution of video popularity.
However, they focus more on the popularity
