Actual Rating Calculation of the Zoom Cloud Meetings App using User Reviews on Google Play Store with Sentiment Annotation of BERT and Hybridization of RNN and LSTM

---

A. Environment

• Google Colaboratory: This project use Google Colaboratory for compute purpose.
• RAM: 12 GB, Disk Space: 80 GB
• Compute Engine: GPU
• Language: Python

B. Datasets

• The datsets used in this work and the saved model data will found in this link.
• Reviews.csv - includes all the raw reviews collected with other information.
• App_Reviews.csv(/.xlsx) - includes only used raw reviews.
• Zoom_Reviews - includes only the zoom app reviews collected.
• /Pickle Data - includes process datasets and models saved data.
• /Pickle Data/Models - includes model saved data for Dataset #1.
• /Pickle Data/Model_2 - includes model saved data for Dataset #2.
• /Pickle Data/Model_3 - includes model saved data for Dataset #3.
• /Pickle Data/Model_4 - includes model saved data for Dataset #4.

C. Codes

• ds1_sentiment_analysis_for_rating_calculation.py
  • Model preparation, training, and testing using DS - 1
• ds2_sentiment_analysis_for_rating_calculation.py
  • Model preparation, training, and testing using DS - 2
• ds3_sentiment_analysis_for_rating_calculation.py
  • Model preparation, training, and testing using DS - 3
• ds4_sentiment_analysis_for_rating_calculation.py
  • Model preparation, training, and testing using DS - 4
• zoom_app_reviews_star_calculation.py
  • Actual rating calculation using models discussed for the Zoom Cloud Meetings App
• all_app_reviews_star_calculation.py
  • Actual rating calculation of All reviews colled using discussed models.
• bert_sentiment_analysis_for_rating_calculations.py
  • Performance Analysis of BERT Base on Pre-Processed Reviews.
• rnn_sentiment_analysis.py
  • Performance Analysis of RNN.
• lstm_sentiment_analysis.py
  • Performance Analysis of LSTM.
• bert_sentiment_analysis
  • Performance Analysis on All Reviews using BERT Base

D. Results and Performance

• Summary of precisions (P), recalls (R), f1-scores (F1), and accuracies (A) of DS - 1, DS - 2, DS - 3, and DS - 4.

		Ratings		DS - 1					DS - 2					DS - 3					DS - 4
				P	R	F1	A		P	R	F1	A		P	R	F1	A		P	R	F1	A
		☆		0.61	0.58	0.59			0.54	0.67	0.60			0.68	0.71	0.69			0.48	0.67	0.56
		☆☆		0.41	0.30	0.35			0.32	0.12	0.18			0.00	0.00	0.00			0.00	0.00	0.00
Our Models		☆☆☆		0.47	0.28	0.35	0.61		0.38	0.29	0.33	0.45		0.33	0.25	0.29	0.71		0.35	0.21	0.27	0.54
		☆☆☆☆		0.43	0.20	0.27			0.37	0.38	0.37			0.38	0.11	0.17			0.41	0.10	0.16
		☆☆☆☆☆		0.67	0.89	0.76			0.49	0.65	0.56			0.76	0.94	0.84			0.60	0.87	0.71
				BERT Base #					BERT Base					RNN					LSTM
		☆		0.81	0.82	0.82			0.53	0.68	0.59			0.40	0.42	0.41			0.38	0.51	0.44
		☆☆		0.51	0.59	0.55			0.30	0.28	0.29			0.00	0.00	0.00			0.00	0.00	0.00
Pretrained Models		☆☆☆		0.33	0.47	0.39	0.67		0.43	0.35	0.38	0.53		0.39	0.35	0.47	0.45		0.34	0.29	0.31	0.47
		☆☆☆☆		0.60	0.61	0.60			0.31	0.40	0.35			0.36	0.06	0.11			0.36	0.11	0.17
		☆☆☆☆☆		0.71	0.67	0.69			0.77	0.67	0.72			0.55	0.83	0.66			0.55	0.83	0.66

# Use pre-processed unbiased dataset based on BERT Base

• Calculated average rating of the Zoom Cloud Meetings App and All Apps with actual average rating.

Model Trained on Dataset	Calculated Average Ratings		Pooled Average Rating		Average Rating on Dataset
	Zoom Cloud Meetings App	All Apps	Zoom Cloud Meetings App	All Apps	Zoom Cloud Meetings App	All Apps
DS - 1	3.70	3.97
DS - 2	3.01	3.03	3.60	3.73	3.08	3.42
DS - 3	3.96	4.21
DS - 4	3.71	3.70

E. Dependencies

• The dataset importation from Google Drive and saving model data into Google Drive required permission of drive.