GitHub - adrianbruenger/twitter_sentiment_lstm: Development of a tweet text sentiment analysis model, using word vectorization and LSTM-Deep learning [TUM-Data Analysis&ML summer 2021]

adrianbruenger / twitter_sentiment_lstm Public

forked from stefanrmmr/kaggle_twitter_airline_sentiment

Notifications You must be signed in to change notification settings
Fork 0
Star 0

Development of a tweet text sentiment analysis model, using word vectorization and LSTM-Deep learning [TUM-Data Analysis&ML summer 2021]

0 stars 1 fork Branches Tags Activity

Star

Notifications

Branches Tags

Name		Name	Last commit message	Last commit date
Latest commit History 65 Commits
.idea		.idea
__pycache__		__pycache__
analysis_plots		analysis_plots
model_data_final		model_data_final
model_data_keras_embedding		model_data_keras_embedding
tweets_data		tweets_data
twitter_acc		twitter_acc
LSTM_model.py		LSTM_model.py
ProjectReport_TwitterLSTM_BruengerRummer_2021.pdf		ProjectReport_TwitterLSTM_BruengerRummer_2021.pdf
README.txt		README.txt
custom_word2vec.py		custom_word2vec.py
embedding_analysis.py		embedding_analysis.py
fundamental_data_analysis.py		fundamental_data_analysis.py
plotting_framework.py		plotting_framework.py
preprocessing.py		preprocessing.py
tweet_sentiment_lstm_keras.py		tweet_sentiment_lstm_keras.py
uc_average_twitter_sentiment.py		uc_average_twitter_sentiment.py

Repository files navigation

@adrianbruenger @stefanrmmr

FINAL PROJECT for "Python for Engineering Data Analysis - from Machine Learning to Visualization"
TUM, summer semester 2021, development of a tweet text sentiment analysis model using an LSTM approach

DATASET & SOURCES
  - [Kaggle] Twitter US Airline Sentiment tweets from February 2015
    https://www.kaggle.com/crowdflower/twitter-airline-sentiment?select=Tweets.csv
 
IMPLEMENTATION & METRICS
  - exploratory data analysis 
  - twitter API integration
  - tweet text preprocessing
  - word embedding/tokenization
    (Word2Vec with continuous Skip-grams)
  - Biderectional LSTM model based
    classifier deep neural network
  - Hyper parameter tuning
  - history accuracy, loss curves
  - metrics precision, recall
  - confusion matrix
  
TWITTER ACCESS DATA
  - In order to access the full real time 
    twitter sentiment analysis application,
    one needs to add their Twitter credentials
    to the twitter_acc\twitter_acc_config.yaml

USE CASES & APPLICATION
  - analyze most recent tweets for
    specific airlines using titter API
    and evaluate average online sentiment