amlan-bora / sentimenta

Sentiment Analysis Model

Sentiment Analysis

This is a sentiment analysis model , implemented on twitter datasets .

In this project, two supervised learning algorithms,Logistic regression and Decision Trees were used on a twitter dataset(testing dataset) containing 17,197 tweets. Out of the two, Logistic Regression turned out to be more efficient than Decision Trees by a sufficient margin(11% - Bag-of-Words, 6%- TF-IDF) on the basis of their F1 scores. However both the feature extraction methods , Bag of Words and TF-IDF turned out to be quite equal in terms of efficiency although TF-IDF had a slightly greater F1 Score for Logistic Regression.

[Code is available in the SAforTwitter.py file and was implemented using Google Colab]

Some of the Data Visualisations done through this project are -

WordCloud of most used Positive Setiment

WordCloud of most used Negative Sentiment

Bar Plot ( Top 20 most frequent words used for hashtag in case of positive tweets)

Bar Plot ( Top 20 most frequent words used for hashtag in case of negative tweets)

Efficiency of ML Models implemented here