Sentiment analysis
Sentiment analysis based on movies reviews.
Table of contents
General info
The main goal of this project is to predict the sentiments of sentences.
We work with IMDB movies reviews dataset. Movies are rated from 1 to 10. We consider only the reviews with rating 1 (the worst ones) and 10 (the best ones).
The assumption is that movies with rating 1 have negative sentiment and movies with rating 10 have a positive sentiment.
Technologies
- Python - version 3.7.3
Libraries
- numpy
- pandas
- matplotlib
- seaborn
- sklearn
- nltk
- scipy
- glob
- re
- cvxopt
- tqdm
Results
- Accuracies below are measured on test data.
- Baseline accuracy: 52%.
- Naive Bayes: 89.95%.
- Logistic Regression: 89.92%.
- SVM: 92.16%.
- Heuristic Naive Bayes: 90.32%.
Status
Project is: finished,
Inspiration
Machine Learning class.
Authors
Created by @TheFebrin, @MatMarkiewicz and @gKlocek