It is a Natural Language Processing work using 2 Machine Learning Algorithms and a Keras Deep Learning model
Its aim is to classify a review as positive, negative or neutral
Lemmatizing, Cleaning, and vectorizing reviews are done by prepare function and nltk library functions
Naive Bayes Classification got %62.1, Logistic Regression Algorithm got %79.7 test accuracy.
Neural Network reaches up to %87 test accuracy which is very nice. Adding layers don't change loss and without initializing first weights with lecun uniform model is limits itself by %80 accuracy. Model may be improved using BERT model and other initializer-regularizer combinations.
Kaggle: https://www.kaggle.com/uurdeep/predict-review-with-ml-and-dl-87-accuracy
Comparisons:
Learning process:
