Please see the following information about the notebooks included in the capstone project:
This poject was done in order to find the dependency of selected model's performance on reviews from different service providers (Yelp, Amazon & IMDb). I wanted to find whether a selected model's performance depend on the company when classifing good and bad reviews. Because, human beings have certain common properties when giving a good comment or a bad comment disreguard of the company they refer to. Main objective of this project is to find how relaiable a traind model on the reviews 1). Traind on mixed reviews, 2). Trained on company specific reviews. More details can be found in Project_Report.pdf frile.
Basic information about some of the codes used in this analysis are given below. [For more information please send me a message : nalinda05kl@gmail.com]
-
Customer_Review_Classification_using_ML.ipynb This notebook shows all the main steps used in this analysis includeing pre-processing and modeling
-
DataVisual.ipynb This includes pre-processing and visualization of raw data.
-
NivGuess_MulNivBays_SGD.ipynb This notebook shows training and testing of MultinomialNB and SGD classifiers.
-
CNN_embed.ipynb This notebook shows the training and testing of different CNN architectures with hidden layers.
-
MNNB.ipynb This note book shows how the optimization of parameters of MultinomialNB was done in order to get the highest accuracy.
-
All the notebooks bellow shows calculation of accuracy and confusion matrix based on MultinomialNB classifier for different training and testing combinations. Train_Combine_Test_Amazon.ipynb Train_Combine_Test_Yelp.ipynb Train_Combine_Test_imdb.ipynb Train_Test_Amazon.ipynb Train_Test_Imdb.ipynb Train_Test_Yelp.ipynb
-
Please make sure to include all .txt data files in the same directory with the above notebooks before running.
-
CNN_with__FreeForm_visualization.ipynb I have added this note book for need of the free form visualization section after the first review of the project.
Thank you. Nalinda Kulathunga.