Twitter Sentiment Analysis

Data can be downloaded here

1. Feature Engineering

• the number of word
• the number of characters
• average word length
• count stopwords
• count hashtags
• count numbers
• count upper case

2. Text Preprocessing

• convert to lower case
• Twitter Handles (@user)
• urls
• punctuations & numbers & Special Characters
• short words that len <= 3 (not applied)
• stop words
• remove the top10 and bottom 10 frequent word (not applied)
• spelling correction (not applied)
• normalize - stem and lemma

3. Feature Extraction

• Bag of words
• TF-IDF
• Word embedding(Word2vec)

4. Split data & Model selection