SENTIMENT ANALYSIS / 30_Days_of_Python #301

[Bharat00007]

Pull Request: Sentiment Analysis Machine Learning Project #301

Overview

This pull request introduces a complete Sentiment Analysis project that utilizes various machine learning models to classify text data, specifically tweets, into positive or negative sentiments. The project covers data preprocessing, feature extraction, model training, and evaluation.

Functionalities

1. Data Loading

• Training and Testing Data:
  • The project begins by loading the training and testing datasets from CSV files. These datasets contain tweets labeled as either positive or negative, which are essential for training and evaluating the models.

2. Data Preprocessing

• Remove User Handles:
  • Twitter handles (e.g., @username) are removed from the tweets using regular expressions.
• Text Cleaning:
  • Punctuations, special characters, and numeric values are removed, leaving only alphabetic characters in the text.
• Remove Short Words:
  • Words with fewer than three characters are removed, as they are unlikely to contribute significantly to sentiment analysis.
• Tokenization and Stemming:
  • The text is tokenized into individual words.
  • Stemming is applied using the Porter Stemmer to reduce words to their root forms (e.g., "running" becomes "run").

3. Data Visualization

• Word Cloud for Sentiments:
  • Word clouds are generated for positive and negative sentiments, providing a visual representation of the most common words in each category.
  • The word clouds are created using a mask image related to Twitter.
• Hashtag Frequency Analysis:
  • Hashtags are extracted from tweets and their frequency is analyzed.
  • The top 20 most frequent hashtags for both positive and negative sentiments are visualized using bar plots.

4. Feature Extraction

• Bag of Words (BOW):
  • The CountVectorizer is used to convert the cleaned tweets into a BOW representation, creating a sparse matrix of word counts.
• TF-IDF:
  • The TfidfVectorizer is applied to transform the text into Term Frequency-Inverse Document Frequency (TF-IDF) features, which provide more refined text representation than raw word counts.

5. Model Training

• Logistic Regression:
  • A Logistic Regression model is trained on both BOW and TF-IDF features to predict sentiment labels.
• XGBoost:
  • An XGBoost classifier, known for its strong performance with structured data, is trained on both BOW and TF-IDF features.
• Decision Tree:
  • A Decision Tree classifier is trained on both BOW and TF-IDF features, providing a non-linear model option for sentiment analysis.

6. Model Evaluation

• F1 Score:
  • The performance of each model is evaluated using the F1 score, a balanced metric that considers both precision and recall.
  • The F1 scores for models trained on BOW and TF-IDF features are displayed, offering insights into their effectiveness.

[github-actions[bot]]

Thank you for submitting your pull request! 🙌 We'll review it as soon as possible. In the meantime, please ensure that your changes align with our [CONTRIBUTING.md]. If there are any specific instructions or feedback regarding your PR, we'll provide them here. Thanks again for your contribution! 😊

[Axwell-2]

Please create an issue of this pull request