Fikir112/fintech-review-analytics

Customer Experience Analytics for Ethiopian Fintech Apps # Fintech Review Analytics ## Overview Customer Experience Analytics for Ethiopian Fintech Apps. This project analyzes Google Play Store reviews for three major Ethiopian banks to deliver actionable insights on user satisfaction, pain points, and feature priorities. ## Banks Analyzed | Bank | App ID | Reviews Collected | |------|--------|------------------| | Commercial Bank of Ethiopia | com.combanketh.mobilebanking | 482 | | Bank of Abyssinia | com.boa.boaMobileBanking | 498 | | Dashen Bank | com.dashen.dashensuperapp | 498 | ## Project Structure fintech-review-analytics/ ├── data/ │ └── raw/ # Scraped and cleaned review data (not tracked by Git) ├── notebooks/ │ ├── task1_scraping.ipynb # Data collection and preprocessing │ ├── task2_sentiment.ipynb # Sentiment and thematic analysis │ ├── task3_database.ipynb # PostgreSQL database storage │ └── task4_visualization.ipynb # Visualizations and insights ├── src/ │ ├── scraper.py # Production scraping and cleaning functions │ ├── sentiment.py # VADER sentiment analysis functions │ └── themes.py # TF-IDF and keyword theme analysis ├── tests/ # Unit tests ├── scripts/ # Standalone scripts ├── .github/workflows/ # CI/CD with GitHub Actions ├── requirements.txt # Python dependencies └── README.md ## Tasks Completed ### Task 1: Data Collection & Preprocessing - Tool: google-play-scraper Python library - Reviews collected: 1,478 (after cleaning from 1,500 raw) - Duplicates removed: 22 - Missing values: 0 - Columns: review, rating, date, bank, source - Error handling: try/except with Python logging throughout ### Task 2: Sentiment & Thematic Analysis - Sentiment tool: VADER (chosen over DistilBERT for speed and suitability) - Scoring: compound score ≥0.05 = positive, ≤-0.05 = negative, else neutral - Results: 61.2% positive, 24.7% neutral, 14.1% negative - Thematic analysis: keyword matching + TF-IDF (scikit-learn …