MOVIE RATING PREDICTION BASED ON TWITTER SENTIMENT ANALYSIS

Journal of Advanced Computing and Communication Technologies (ISSN: 2347 - 2804)
Volume No 5 Issue No. 1. February 2017
6
MOVIE RATING PREDICTION BASED ON TWITTER
SENTIMENT ANALYSIS
By
1
Mr.Abhishek Kesharwani, 2
Mr. Rakesh Bharti
M.Tech Scholar, UIT, Dr. A. P. J. Abdul Kalam Technical University, Uttar Pradesh, Lucknow, India
Asst Prof., UIT, Dr. A. P. J. Abdul Kalam Technical University, Uttar Pradesh, Lucknow, India
abhishekkesharwani689@gmail.com,goswami.rakesh@gmail.com
ABSTRACT
With microblogging platforms such as Twitter generating
huge amounts of textual data every day, the possibilities of
knowledge discovery through Twitter data becomes
increasingly relevant. Similar to the public voting mechanism
on websites such as the Internet Movie Database (IMDb) that
aggregates movies ratings, Twitter content contains
reflections of public opinion about movies. This study aims to
explore the use of Twitter content as textual data for
predicting the movie rating. In this study, we extract number
of tweets and compiled to predict the rating scores of newly
released movies. Predictions were done with the algorithms,
exploring the tweet polarity. In addition, this study explores
the use of several different kinds of tweet classification
Algorithm and movie rating algorithm. Results show that
movie rating developed by our application is compared to
IMDB and Rotten Tomatoes.
General Terms
Movie rating prediction, tweet polarity
Keywords
Twitter sentiment analysis, comparison of twitter rating from
IMDB and Rotten Tomatoes, tweet classification algorithm,
movie rating algorithm.
1. INTRODUCTION
The task of movie rating prediction by extracting the tweets
and hashtags related to particular movie. Twitter data can be
accessed through the public API provided by the Twitter.
These APIs can be accessed only by authentication requests,
which must be signed with valid login ID and password.
Twitter provides authentication keys for extractions of the
tweets. Tweet extracted from twitter API is collected in
MySql where the tweet is having a unique tweet id, twitter id
and tweet date. Movie name is added by the user and the
release date of the movie is also added by the user. We have
to add the twitter Id of the movie from the twitter and all the
tweets are extracted from the twitter. We have to update the
tweets of the movie by establishing the connection with
twitter from our application, and then all the recent tweets are
added to our database. We have also calculated the popularity
of the movie at the time we extract the tweets. Each time we
want to predict the rating or calculate the popularity of movie
labelling of every tweet is also calculated. Apart from the
tweets obtained from Twitter, the application also calculates
the sentiment associated with the tweet. The application also
stores the sentiment associated with the tweet in the database.
Sentiment of a tweet is categorized as positive, negative,
neutral and irrelevant. In this application we developed a
module that is used to create bag a words from the tweets of
old movies. We create two bags of words positive and
negative .This is the backbone of our application. Each tweet
that we extract from the twitter should be categorised as
positive or negative tweet. We have a module update positive
and negative words in my application which is used for
updating positive and negative bag of words. For calculating
the movie rating we ignored neutral and irrelevant tweets as
these are not useful for any type of information for movie
review. For movie rating prediction we have designed a
module by which you can select any movie and all the tweets
related to that movie is loaded in the algorithm with hashtags
of the movie is also added so that all the review is added for
the particular movie and more precise and accurate rating is
calculated. When we search for a particular movie which is
twitter id of the movie then rating out of 10 is calculated and
change to format of 5 star rating with the help of our
algorithm we design for movie rating. The performance of this
application is evaluated by comparing its results with results
from popular movie rating websites like IMDB and Rotten
Tomatoes. The rating calculated for various movies released
by fetching tweets related to the movie is stored in our
application and real time rating of the movie is calculated and
is represented in tabular form and graphically. The ratings
collected show that the Twitter rating application is following
similar trends as shown in IMDB and Rotten Tomatoes.
Ratings from Twitter application are observed to be of a lower
value. This can be attributed to the fact that the number of
users rating a movie on the other websites is almost 100x
times than what the Twitter application is using. A small
experiment was conducted with varying number of tweets
used by Twitter application to confirm this theory.
2. RELATED WORK
The topic of using social media to predict the future becomes
very popular in recent years. Different work has been already
done using twitter content for predicting the sentiment of
tweets. Movie sentiment analysis based on public tweets [1]
,In this paper they introduced a special approach to
understanding the tongue and frequency of words from one

7
sentiment category enabling a much better sentiment
classification compared to the normal machine learning
techniques. They additionally introduce an added sentiment
category - the neutral category. In their analysis they use the
Python programming language with the NLTK library and
compare so obtained results with the normal machine
learning. Predicting Ratings for New Movie Releases from
Twitter Content [5] that is textual knowledge from Twitter
will be seen as an in depth supply of data relating to a
particularly broad form of subjects. With a lot of users
actively expressing themselves on-line, a large quantity of
information is generated daily. Since this knowledge for an
oversized half consists of human expressions, Twitter
knowledge will be seen as a valuable assortment of human
opinion or sentiment, which might be mechanically extracted
with comparatively high accuracy.
3. METHODOLOGY
The overall system can be designed in following phases
i. Tweet Collection
ii. Tweet Classification
iii. Rating Movies.
3.1 Tweet Collection
Twitter data can be accessed through the public API provided
by the Twitter. These APIs can be accessed only by
authentication requests, which must be signed with valid login
ID and password. Twitter provides authentication keys for
extractions of the tweets. We have to follow some steps to
create Authentication keys.
i. Create application on twitter.
ii. Manage Application
iii. Change the permissions to read and write.
iv. Retrieve Authentication keys.
First we create application on twitter by signing in to
https://apps.twitter.com/app/new. Second Step we have to
manage Application. In the third step change the application
permissions to read and write. After the completion of the
Application ,we have the following unique keys that is
required to fetch tweets from twitter
a. Consumer key
b. Consumer Secret key
c. Access token
d. Access token secret
Tweet Extracted from twitter having complete information
like date of tweet, tweet ID, user ID, retweet count etc.We
will use only tweet date, tweet ID and tweet. We will add two
more columns release date and added by to have some extra
information about the analysis of movie. We have integrated
twitter API into our application so as to fetch all the tweets
related to a particular movie and all the news and comments
related to a particular movie. There are so many limitation
related to this API as this will extract limited tweets
approximately hundred tweets in one time. But the tweets are
collected into the database as we have MySql database in our
back end. We have to update the tweets of the movie by
establishing the connection with twitter from our application,
and then all the recent tweets are added to our database.
3.2 Tweet Classification
In this module I label every tweet as positive, negative,
neutral and irrelevant. Each time we want to predict the rating
or calculate the popularity of movie and labeling of every
tweet is also calculated. Apart from the tweets obtained from
Twitter, the application also calculates the sentiment
associated with the tweet. In this we tokenize the tweet into
different tokens separated by space and compare each token
with our predefined set of positive and negative bag of words.
After the comparison of tokens we find the total number of
positive and negative tokens in the tweet. Count the total
number of positive and negative tokens in the tweet and label
them as p and n respectively. Calculate the value of ratio as
total number of positive tokens to the total number of positive
and negative token.
Table 1 Table for tweet labeling
Ratio Tweet Label
ratio>0.5 Positive
ratio=0.5 Neutral
ratio<0.5 Negative
ratio=0/0 (p==0 && n==0) Irrelevant
3.2.1 Bag of words
In this application we developed a module that is used to
create bag a words from the tweets of old movies. We create
two bags of words positive and negative .These are the
backbone of our application. Each tweet that we extract from
the twitter should be categorized as positive or negative tweet.
We have a module update positive and negative words in my
application. In this module if we want to add any positive or
negative word ,if that word already exist in the database of
bag of words then it is not updated.
3.3 Algorithm for Tweet Classification
1. Extract the tweet from twitter using Twitter API.
2. Store the tweet in the database. Each tweet has a unique id
and date of tweet which is also store in the database.
3. For each tweet we have to specify the tweet label. Tweet
labels are positive, negative, neutral and irrelevant.
4. Calculate the value of p and n i.e. total number of positive
and negative words obtained by comparing each word
stored in our manually collected positive and negative
collection of bag of word.
5. Classify the tweet label of each tweet using the following
formula
ratio= p/ (p + n)
If (ratio>0.5)
{
Tweet label="positive";
}
If (ratio==0.5)
{
Tweet label="neutral";
}
If (ratio<0.5)
{

8
Tweet label="negative";
}
If (p==0 and n==0)
{
Tweet label="irrelevant";
}
6. Here p and n are total number of positive and negative
words obtained by comparing each word stored in our
manually collected positive and negative collection of
word set.
3.4 Rating Movie
Rating is calculated by the total number of positive tweets
over total tweets. This was designed that the rating is as
correct because it will get the most recent views about the
particular movie and twitter chat about the popularity and
criticism of the movie
3.5 Algorithm for Rating Movie
1. Update the tweets and hashtag of a movie with twitter id
and store it into database.
2. Select all the tweets and hashtags from the database with
same twitter id and apply tweet classification algorithm.
3. Classify all the tweets and store the tweets with tweet
labels as positive, negative, neutral and irrelevant.
4. Ignore neutral and irrelevant tweets. Neutral tweet does
not specify any positive or negative sentiment as they
create a situation of ambiguity and irrelevant tweets do
not specify any sentiment.
5. Calculate the value of rating as
rating = ((total positive tweets) / (total positive tweets
+total negative tweets))*10)
6. if (rating<2.0 and rating >=0)
{
Movie rating = 1 star
}
if ( rating>=2 and rating<3 )
{
}
if (rating>=3 and rating<4 )
{
Movie rating = 2.5 star
}
if (rating>=4 and rating<5 )
{
}
If (rating>=5 and rating<6)
{
}
If (rating>=6 and rating<8 )
{
}
If (rating>=8 and rating<9)
{
}
If (rating>=9 and rating <=10)
{
}
4. EXPERIMENTAL RESULTS AND
PERFORMANCE ANALYSIS
The performance of this application is evaluated by
comparing its results with results from popular movie rating
websites like IMDB and Rotten Tomatoes. IMDB is internet
information of data associated with films, television,
programs etc. the location permits registered users to rate any
film on a scale of one to ten, except for writing reviews
regarding it. The location displays a weighted mean of user
ratings and displays it next to the movie title. This website has
6.05 Daily Page views per visitor [7] and a median of fifteen
million individuals visit the web site per month. Rotten
Tomatoes may be a web site dedicated to film reviews and
news. It offers 2 sorts of scores for movies – Tomato meter
critic mixture score and Audience score. The Critic mixture
score reflects reviews and ratings from varied newspaper
writers or from those that belong to film critic associations.
The Audience score is calculated supported user’s reviews
and ratings. For the aim of confirmatory this application,
solely Audience score is taken into account. Registration is
free however the location requests permission to look at user’s
social network profiles. It gets 3.60 Daily Page views per
visitor and a median of thirteen million individuals visit the
web site per month. [8].The rating calculated for various
movies released by fetching tweets related to the movie is
stored in our application and real time rating of the movie is
calculated and is represented in tabular form and graphically
as shown in table 2 and figure 1.
Table 2 Twitter Rating vs. IMDB and Rotten tomatoes
Movie
Name
Twitter
Rating IMDB
Rotten
Tomatoes
Ae dil hai
mushkil 2 3 2.9
Shivaay 3.5 3.5 3.8
Force2 3 3 3.2
Pink 3 4 4.2
Raees 2 4 3.9

9
Kaabil 4 4 4
Dangal 3.5 4.5 4.6
Kahaani2 4 3.5 3.7
Bahubali 2.5 4 4.3
As shown in the Figure 1 twitter rating for the movie “Ae Dil
hai Mushkil” is 2 while that IMDB is 3 and rotten tomatoes
rated it 2.9.For the movie “Kaabil” we predict the same rating
as predicted by the IMDB and Rotten Tomatoes. For the
movie “kahaani2” we predict rating as 4 but IMDB rated as
3.5 and rotten tomatoes rated 3.7.
Fig 1: Twitter Rating versus IMDB and Rotten Tomatoes
Table 3 Twitter Reviews versus IMDB and Rotten
Tomatoes Reviews
Movie
Name
Twitter
Reviews
IMDB
Reviews
Rotten
Tomatoes
Reviews
Ae dil hai
mushkil 1375 12644 921
Shivaay 392 6341 154
Force2 233 2245 50
Pink 199 18102 302
Raees 431 14390 332
Kaabil 1200 7462 131
Dangal 193 30720 1433
Kahaani2 200 1719 204
Bahubali 431 67698 1743
As show in Table 3 of Twitter Reviews and IMDB and Rotten
Tomatoes reviews, the number of reviews are very large for
IMDB as compared to twitter reviews for the movie “Ae dil

10
hai mushkil” total number of tweets we collected are 1375
while that of IMDB is 12644 and that of Rotten Tomatoes we
obtained the ratings as 2 for twitter, 3 for IMDB and 2.9 for
Rotten Tomatoes. For the movie “Kaabil” total number of
tweets we collected are 1200 while that of IMDB is 7462 and
that of Rotten Tomatoes 131 we obtained the ratings as 4 for
all are same. The result shows that if we obtain huge amount
of data from twitter than we calculate more accurate rating by
our algorithm as this application works more accurately with
huge amount of data, but currently the application can retrieve
close to 100 tweets per API call, depending on how recently
the movie was tweeted about. It has to be stressed that this is
not enough data to determine the opinion of the Crowd for
rating a movie. This is especially the case when dealing with
Twitter data retrieval based on keyword search. The search
results invariably include advertisements, retweets, and spam.
These have to be filtered out or categorized separately in order
to not interfere with the sentiment classification. Since Twitter
has a rate limit on data available for use, the workaround for
this issue in the application is to store the tweets every time it
is retrieved, and calculate a new rating when the movie name
is searched again. These results in an increase of data used to
review movies, every time the application is used.
5. CONCLUSION
Twitter data effectively manages to capture the opinions and
emotions of the crowd and Twitter APIs make it fairly easy to
gather this information and analyze it. This desktop
application indeed manages to use this massive amount of
data to provide a meaningful and useful result due to the
speed limit introduced by Twitter, this is often presently an
educational implementation of the thought to use twitter
knowledge for rating a movie, In future if the limit on the
information of twitter is removed then this application has
excellent result and may be used for thus several merchandise
review and its quality. If all the tweets containing the search
string for a movie name will be captured and analyzed, a lot
of precise conclusion will be drawn.
The training data used by the classification algorithm are very
limited for tweets per category. The results have been
encouraging with the small set; I expect the result to be even
more impressive with a larger selection of natural language
processing in the training data. For some movies we predicted
the exact rating but in some cases we are not getting exact
result as we have some less data for that movie.
As the application is used frequently, the dataset grows with
it. This results in the rating of a movie always being up to date
with public opinion. The best use case is to search for recent
movies, simply because that is when the crowd seems to be
tweeting most about the movies then more accurate rating will
be predicted.
6. FUTURE WORK
In future enhancements, more categories can be introduced to
classify tweets – extremely positive, mildly positive,
extremely negative, mildly negative, neutral, and irrelevant.
This can be used to improve the rating formula to make it
more accurate. A weight can be associated with each category
and then calculate the average.
More number of classes implies increase in number of
attributes for the classifier to compare the input with. This
should show some difference in the outcome of the classifier
and hence difference in the rating shown by the application.
We only consider the data from twitter but in future data from
other social media like facebook, YouTube and other blogger
comments is also taken into considerations for more accurate
rating and with real opinions from all the social media into
single place.
7. REFERENCES
[1] Aljaˇz Blatnik, Kaja Jarm, Marko Meˇza, Movie sentiment
analysis based on public tweets, Faculty of Electrical
Engineering, University of Ljubljana, Trˇzaˇska 25, 1000
Ljubljana, Slovenia, 81(4): 160–166, 2014 ORIGINAL
SCIENTIFIC PAPER.
[2] Snehal. A. Mulay, Shrijeet J Joshi, Mohit R Shaha,
Hrishikesh V Vibhute, Mahesh P Panaskar, Sentiment
Analysis and Opinion Mining With Social Networking for
Predicting Box Office Collection of Movie, International
Journal of Emerging Research in Management &Technology
ISSN: 2278-9359 (Volume-5, Issue-1)
[3] Ladislav Peska, Peter Vojtas, Hybrid Biased k-NN to
Predict Movie Tweets Popularity, Faculty of Mathematics and
Physics Charles University in Prague Malostranske namesti
25, Prague, Czech Republic.
[4] Vasu Jain, Prediction of Movie Success using Sentiment
Analysis of Tweets Department of Computer Science,
University of Southern California,
The International Journal of Soft Computing and Software
Engineering [JSCSE], Vol. 3, No. 3, Special Issue.
[5] Wernard Schmit, Sander Wubben, Predicting Ratings for
New Movie Releases from Twitter Content, Proceedings of
the 6th Workshop on Computational Approaches to
Subjectivity, Sentiment and Social Media Analysis (WASSA
2015), pages 122–126,Lisboa, Portugal, 17 September, 2015.
c 2015 Association for Computational Linguistics.
[6] IMDb.com, Inc. Opening this week.
http://www.imdb.com/ (accessed December 2016).
[7] Alexa. Statistics for IMDB, available at
http://www.alexa.com/siteinfo/imdb.com
[8] Alexa. Statistics for Rotten Tomatoes, available at
http://www.alexa.com/siteinfo/rottentomatoes.com
[9] Twitter, Inc. Twitter Help Center - Using hashtags on
Twitter. https://support.twitter.com/entries/49309 (accessed
February 2017)
[10] Twitter, Inc. Discover Twitter - What is Twitter and how
to use it. https://discover.twitter.com (accessed February
2017).
[11] Twitter, Inc. About. https://about.twitter.com/company
(accessed December, 2016).
[12] Twitter, Inc. Twitter developers - Documentation.
https://dev.twitter.com/ (December, 2016)

MOVIE RATING PREDICTION BASED ON TWITTER SENTIMENT ANALYSIS

More Related Content

What's hot (20)

Similar to MOVIE RATING PREDICTION BASED ON TWITTER SENTIMENT ANALYSIS (20)

More from Editor Jacotech (20)

Recently uploaded (20)

MOVIE RATING PREDICTION BASED ON TWITTER SENTIMENT ANALYSIS