b-ezawit/FinTech-Review-Analysis

An NLP-powered analytics pipeline that analyzes Ethiopian fintech app reviews to uncover customer sentiment, key issues, and product improvement opportunities. # Fintech Review Analytics: Customer Experience Analysis for Ethiopian Banking Apps ## Overview This project analyzes customer reviews from Ethiopian banking mobile applications to uncover user satisfaction drivers, common complaints, and product improvement opportunities. The project was completed as part of the **10 Academy Artificial Intelligence Mastery Program – Week 2 Challenge**. The analysis covers three Ethiopian banks: - Commercial Bank of Ethiopia (CBE) - Bank of Abyssinia (BOA) - Dashen Bank The pipeline transforms raw Google Play Store reviews into actionable insights using: - Web scraping - Data preprocessing - Sentiment analysis - Thematic analysis - PostgreSQL database engineering - Data visualization --- # Business Problem Customer reviews contain valuable information about application performance, usability, and user expectations. However, this feedback is mostly unstructured and difficult to analyze manually. This project helps answer: - What do users like or dislike about Ethiopian banking apps? - What issues affect customer satisfaction? - Which features should banks prioritize improving? --- # Project Workflow ``` Google Play Store Reviews | ↓ Data Collection | ↓ Data Cleaning & Preprocessing | ↓ Sentiment Analysis | ↓ Theme Extraction | ↓ PostgreSQL Storage | ↓ Visualization & Recommendations ``` --- # Technologies Used ## Data Collection - Python - google-play-scraper ## Data Analysis - pandas - NumPy - Scikit-learn - spaCy - NLTK ## NLP - DistilBERT Sentiment Analysis Model - TF-IDF Keyword Extraction ## Database - PostgreSQL - psycopg2 / SQLAlchemy ## Visualization - Matplotlib - Seaborn ## Testing & Automation - pytest - GitHub Actions --- # Dataset Reviews were collected from the Google Play Store for: | Bank | Reviews | |---|---| | Commercial Bank of Ethiopia | 400+ | | Bank of Abyssinia | 400+ | | Dashen Bank | 400+ | Collected fields: - Review text - Rating - Review date - Bank name - Source platform --- …