Scribd
Upload
34 views19 pages

Sentiment Analysis IMDB Review - Presentation

Uploaded by

varun190104
Copyright
© © All Rights Reserved
We take content rights seriously. If you suspect this is your content, claim it here.
Available Formats
Download as PDF, TXT or read online on Scribd
34 views19 pages

Sentiment Analysis IMDB Review - Presentation

Uploaded by

varun190104
Copyright
© © All Rights Reserved
We take content rights seriously. If you suspect this is your content, claim it here.
Available Formats
Download as PDF, TXT or read online on Scribd
You are on page 1/ 19

Sentiment Analysis

of IMDB Movie
Reviews

BINGHAMTON UNIVERSITY
STATE UNIVERSITY OF NEW YORK
Team Members -
SPRING 2020
CS-580L-01

- Chaitanya Kulkarni
B-Num: B00814455
SUNY Binghamton
[email protected]

- Abhimanyu Singh
B-Num: B00813542
SUNY Binghamton
[email protected]

- Necati Anil Ayan


B-Num: B00777933
SUNY Binghamton
[email protected]
1. Motivation
- Most convenient source of entertainment
- Often confused about whether to watch that
particular movie or not
- We check websites for ratings and reviews
- Most of them show rating based on the stars
given
- But no method to know about the success of
the movies based on the reviews and
comments
2. Introduction
- To determine the success or failure based on
the reviews, we analyze the text
- Interpret and classify the emotions within
textual data
- Focuses on polarity - Positive reviews or
Negative reviews
- We used 3 classifiers -
● Logistic Regression
● Naive Bayes
● Support Vector Machine (SVM)
3. Dataset Info
- The dataset can be found at -
Large Movie Review Dataset v1.0

- Dataset provided by Stanford


- 50,000 movie reviews from IMDB website
- 25K for training data & 25K for testing data
4. Data Preprocessing
- Removing HTML tags
- Remove special characters and stopwords
- Lemmatization
- Tokenization
5. Feature Extraction
- Is used to convert feature (words in this case)
into some form.
Bag of Words Approach
● Term Frequency
● CountVectorizer with scikit-learn
6. Classification Models
- Classifiers -
● Logistic Regression
● Naive Bayes
● Support Vector Machine (SVM)
A. Logistic Regression
- Discriminative Model
- Logistic Function (Mostly Sigmoid)
- Pretty good for binary output model (positive or
negative)
- Not good for non-linear solutions
B. Naive Bayes
- Generative Model
- Assumes all features are conditionally
independent
- Based on Bayes Theorem

- May be absurd for real-life cases since


predictors are usually dependent.
C. Support Vector Machine (SVM)
- Objective is to define hyperplane in
N-dimensional space to classify data points.
- Goal is the maximizing margin (between two
data points)
- Also good for non-linear solutions (Kernel trick)
- Very suitable for text classification
7. Result Evaluation
● ACCURACY
● A result evaluation measure
● It is the ratio of correctly predicted
observations to the total number of
observations
● Higher the accuracy better the
model
7. Result Evaluation(cont.)
● LOGISTIC REGRESSION
● Accuracy: 86.89%
● SUPPORT VECTOR MACHINE
● Accuracy: 85.29%
● NAIVE BAYES
● Accuracy: 85.48%
8. Conclusion
- Highest accuracy obtained on Logistic
Regression model.
- LR performs better than Naive Bayes
- LR outperforms the SVM, which is known to be
the best choice for textual data.
9. Our Learning
● Before handling the data we need to clean the data. This is called data
preprocessing. Necessary to clear missing, noisy and inconsistent data which
might affect accuracy.
● Semintent Analysis is extraction of the emotions from within the textual data. It is a
technique based upon Natural Language Processing.
● Logistic regression and support vector machines are closely linked. Both can be
viewed as taking a probabilistic model and minimizing some cost associated with
misclassification based on the likelihood ratio.
● Naive Bayes classifier is the generative model. Naive Bayes also assumes that the
features are conditionally independent.
● Both Naive Bayes and Logistic regression are linear classifiers, Logistic Regression
makes a prediction for the probability using a direct functional form whereas
Naive Bayes figures out how the data was generated given the results.
● Thus, when the training size reaches infinity then discriminative model logistic
regression performs better than the generative model Naive Bayes.
● Hence in our project we get more accuracy for Logistic Regression than the other
two classifiers.

- Chaitanya Kulkarni
9. Our Learning
● Understood that main difficulty is to make dataset ready for training (Data
Preprocessing).
● Understood the differences between generative and discriminative
models.
● When dataset size is very large like infinite discriminative models are
preferable. However, Generative models can reach its asymptotic faster
since it needs fewer training set to do that.
● As known that SVM is the one the best classifiers. However, it performs
poorly in our dataset. Main reason of this may be we have ignored
hyperparameter tuning. If we consider hyperparameter tuning, accuracies
for SVM and LR may increase but Naive Bayes accuracy may remain stable.

- Necati A Ayan
9. Our Learning
● Got to learn that, constructing and training a model is only a small part of a
machine learning project, major task lies in Data Pre-processing
● Learned various ways to clean the textual data and why is it necessary, to
make it suitable to feed to our learning model.
● Learned to apply what was taught in class in our project, like which models
to choose for a specific task, which in our case was text classification.
● Learned how to improve the accuracies of the models by employing
various optimization techniques like hyperparameter tuning
● Learned the actual difference between probabilistic and Binary Classifiers
● Got to know that, though some models are considered better for some
tasks, but other models can outperform them in some data-sets, which
happened in our case.

- Abhimanyu Singh
10. References
[1]. MaisYasen, Sara Tedmori. “Movies Reviews Sentiment Analysis and
Classification”. IEEE Jordon International Joint Conference on Electrical
Engineering and Information Technology (JEEIT). 978-1-5386-7942-5.
[2]. Tirath Prasad Sahu, Sanjeev Ahuja. “Sentiment Analysis of movie reviews: A
study on feature selection and classification algorithms”. International Conference on
Microelectronics, Computing, and Communication (MicroCom).978-1-4673-6621-2.
[3]. Wijayanto, Unggul and Sarno, Ritanarto. “An Experimental Study of Supervised
Sentiment Analysis Using Gaussian Naïve Bayes”.
476-481.10.1109/ISEMANTIC.2018.8549788.
[4]. Tejaswini M. Untawale, G. Choudhari. “Implementation of Sentiment
Classification of Movie Reviews by Supervised Machine Learning Approaches”.
978-1-5386-7808-4.
Thank You

You might also like