nblomerus/South-African-Bank-Sentiment-Analysis

Investigation of the sentiment of the top South African banks # Twitter Analysis of Popular South African Banks # Overview This is project is aimed at analyzing public opinion towards various South African banks. The project uses natural language processing techniques to process large amounts of data. The analysis will provide insights into customer satisfaction, areas of improvement and general public perception of the banks. The results of this project can be useful for the banks to improve their services and reputation in the South African market. # Dataset The dataset for this project was collected using the sntwitter module. A total of 5,591,765 tweets were scraped using the following search dictionary: ``` bank_dict = {'fnb': ['fnb', 'FNBSA', 'fnbSouthAfrica'], 'absa': ['absa', 'absaSA', 'ABSASouthAfrica'], 'nedbank': ['nedbank', 'NEDBANKSA', 'nedbankSouthAfrica'], 'capitec': ['capitec', 'CapitecBank', 'capitecSA', 'capitecbankSA'], 'standard_bank': ['standard bank','standardbank', 'StandardbankSA', 'standardbankZA', 'standardbankSouthAfrica']} ``` The tweets were collected between 1 January 2006 to 1 January 2023. Each tweet contains the following information: Tweet text Date and time of tweet Tweet author Number of retweets Number of likes All data sources have been upload to Kaggle here. Data has been stored in both parquet and excel file formats. You can use the df_to_excel notebook to convert the parquet files to excel sheets. Note: The data collected may contain inaccuracies or unrepresentative samples, as it is based on public tweets and may not accurately reflect the overall sentiment towards the banks. # NLP models For the natural language processing (NLP) tasks in this project, the tweetnlp module developed by CardiffNLP (available at github.com) was used. The module provides several pre-trained models for sentiment analysis, hate speech detection, offensive language detection and topic detection. Sentiment Analysis Model: The sentiment analysis model is used to classify the sen …