jossieT/ethiopian-mobile-banking-cx-insights

Data-driven customer experience insights for Ethiopian mobile banking platforms using Google Play Store review scraping and analysis. # Ethiopian Mobile Banking CX Insights An analysis pipeline for collecting and preprocessing Google Play reviews for three Ethiopian banks: **CBE**, **BOA**, and **Dashen**. This repository contains scraping code, a preprocessing pipeline, and an EDA notebook with visualizations to help measure user experience and service quality over time. **Status:** Task 1 — Data Collection & Preprocessing --- **Project Goals** - Collect user reviews and metadata from Google Play for three target bank apps. - Clean and normalize the data to a reliable CSV for analysis. - Clean duplicate data - Clean non-english like Amharic reviews - Remove emoji from the review text - Provide exploratory visualizations showing rating trends, review volumes, and text statistics. --- **Minimum Deliverables** - Scraped raw data: at least **400 reviews per bank** (≈1,200 total). - Preprocessing script: `Scripts/preprocessing.py` (handles missing values, dedupe, language filter, date normalization). - Processed CSV: `Data/processed/reviews_processed.csv`. - Analysis notebook: `Scripts/preprocessing_EDA.ipynb` (plots + EDA). --- **Repository Structure (key files)** - `Scripts/` - `scraper.py` — scraper using `google-play-scraper` to collect reviews and metadata - `preprocessing.py` — preprocessing pipeline (load → clean → dedupe → filter → normalize → save) - `preprocessing_EDA.ipynb` — notebook with EDA and plots - `config.py` — app IDs, bank name mapping, data paths, scraping configuration - `Data/` - `raw/` — raw output from the scraper - `processed/` — cleaned datasets - `.env` — optional local overrides (not committed) - `.gitignore` — excludes `data/`, `.venv`, `.env`, and caches - `requirements.txt` — required Python packages --- **How it works — High Level** 1. Scrape - `scraper.py` fetches reviews, ratings, dates, reviewer name, thumbs-up counts and other metadata for each app. - Outputs combined raw CSV to `Data/raw/reviews_raw.csv`. 2. Preprocess - Handle missing critical f …