-
The project begins in the scraped_reviews folder for webscraping reviews for;
-
The scraped data is saved in one database using PostgreSQL named amazon_sentiment with respective tables namely; led_wireless_mouse, keyboard, speaker, headset. The data is then accessed and exported to the data_analysis folder for data analysis.
-
Within the data analysis folder, pre-processing, EDA and Clustering analysis were carried out for modeling purposes within the model folder.
- Two approaches will be taken into account using a cleaned dataset without spelling correctiion and one with spelling correction and see which result performs better (cleaned_reviews & correct_spelling_reviews).
- Moreover, the WordCloud images can be seen in the img folder.
-
The model folder has two approches ml_algo and deep_learning for results comparison.
NB: Classes 0, 1 & 2 represent negative reviews (-1), neutral (0)reviews & positive reviews(1) respectively. Please refer to ml_algo/README.md for experimental results on deep learning vs ML_algo