KokebBenti/Customer-Experience-Analytics

This project centers on analyzing customer satisfaction with mobile banking apps by collecting and processing user reviews from the Google Play Store for three Ethiopian banks: Commercial Bank of Ethiopia (CBE), Bank of Abyssinia (BOA), and Dashen Bank. We'll scrape app reviews, analyze sentiments and themes, and visualize insights. ## Customer Experience Analytics for Fintech Apps: A Real-World Data Engineering Challenge: Scraping, Analyzing, and Visualizing Google Play Store Reviews **Summary** This repository contains the analysis of the reviews for the mobile applications of 3 banks from Google Play to figure out the ways of improving these apps. It conatins - Web scrapping - Using NLP for sentiment analysis - Key word extraction - Database engineering **Introduction** This week, we analyzed how the customers of three banks are saying about the banks’ mobile applications. We are representing Omega consulting and trying to help the banks improve their mobile apps to enhance customer retention and satisfaction. The project was designed to explore web scraping, thematic NLP analysis and data engineering. Methodology We have three banks: Commercial Bank of Ethiopia (CBE), Bank of Abyssinia (BOA), and Dashen Bank. To complete the project and find the necessary answers, we followed the following steps. **1. Web scrapping** • We collected the user reviews from Google Play using the Python library Google-Play-Scraper. We collected dates, reviews and ratings for each app using the company Application ID. • We then proceeded to clean the data. This includes  Removing duplicate rows in the tables.  Deciding what to do with the missing data (in this case drop the rows is it is in an important column)  Making sure each column has the correct data type.  Saving the cleaned data to CSV. The cleaned data looks as below for Bank of Abyssinia. **2. Sentiment and Thematic Analysis** To understand what the overall theme in the reviews and find out the major pain points that customers face, we use sentiment analysis. This includes: • Using distilbert-base-uncased-finetuned-sst-2-english and textblob to compute sentiment scores for each review. We can compare the two models and notice they have mostly similar results as seen below. • Conduct thematic analysis to summarize user feedback into actiona …