Based on: Sentiment Analysis of IMDB Movie Reviews

Created for a group assignment as part of a Machine learning and Data mining course

• Run pip install -r requirements.txt to install all requirements. Some nltk parts might need to be installed manually; check the terminal.
• Run main.py
• Select method, sample size, model
• LIME explanation gets created and shown in the notebook as html

• Logistic Regression (Bag of Words) Accuracy: 0.8738
• Random Forest (Bag of Words) Accuracy: 0.8495
• SVM (Bag of Words) Accuracy: 0.8719

• Logistic Regression (TF-IDF) Accuracy: 0.8934
• Random Forest (TF-IDF) Accuracy: 0.8513
• SVM (TF-IDF) Accuracy: 0.8991

Logistic Regression Accuracy: 0.8443 Logistic Regression Precision: 0.8420425193721438 Logistic Regression Recall: 0.8476

Random Forest Accuracy: 0.805 Random Forest Precision: 0.801621835443038 Random Forest Recall: 0.8106

SVM Accuracy: 0.8571 SVM Precision: 0.8534943575529598 SVM Recall: 0.8622

• Sentiment Classification Accuracy: 0.8576

                precision    recall  f1-score   support
      negative       0.86      0.86      0.86      4942
      positive       0.86      0.86      0.86      5058
  
      accuracy                           0.86     10000
     macro avg       0.86      0.86      0.86     10000
  weighted avg       0.86      0.86      0.86     10000
  

              precision    recall  f1-score   support

    negative       0.00      0.00      0.00      4987
    positive       0.50      1.00      0.67      5013

    accuracy                           0.50     10000
   macro avg       0.25      0.50      0.33     10000
weighted avg       0.25      0.50      0.33     10000

• SVM always seems to take the longest out of the options, while logistic regression is decently fast and provides basically the same accuracy.
• Random forest performs the worst.

• Clustering seems to behave weirdly for now. Unsure if this is normal or if the code is wrong.

• Improve Preprocessing (Add Lemmatization and see if stemming/lemmatization/stemming + lematization works best) - IN PROGRESS!

• Check what is NER and implement it - IN PROGRESS!

• Word clouds (for positive and negative decisions)
• Clustering
• Classification including confusion matrix - T.SNE