How souley225/customer_segmentation_project works
Project type and purpose
• This is a full-stack, modular Data Science application in Python.
• Purpose: run a reproducible RFM segmentation + Market-Basket Analysis pipeline and expose results in an interactive Streamlit dashboard for customer segmentation and product cross-sell/upsell recommendations.
File structure
─ Top-level:
app.py → entry point redirecting to the Streamlit app
config/ → YAML config (paths, thresholds, parameters)
models/ → serialized outputs (rfm_segments.csv, association_rules.csv, metadata.json)
src/ → core pipeline modules (data_cleaning, rfm_segmentation, basket_analysis, recommendation, visualization, utils)
requirements.txt → list of Python dependencies
─ app/
├─ streamlit_app.py → UI layer
├─ components/ → reusable UI widgets
└─ assets/ → static assets (images, CSS)
─ Note: There is no external database yet, data is read from files (raw → interim → processed).
README for additional insights
• Describes four main stages:
1. Data cleaning
2. RFM segmentation
3. Market Basket Analysis (Apriori)
4. Streamlit Dashboard
• Specifies reproducibility, modularity, and possible future FastAPI or SQL store additions.
a. Main components
• Configuration loader
• Data pipeline:
– Data Cleaning module
– RFM Segmentation module
– Basket Analysis module
– Recommendation engine
– Visualization utilities
• Storage:
– Raw / interim / processed data folders
– models/ output store (CSV, JSON)
• Frontend:
– Streamlit app (app/streamlit_app.py + components)
• (Future) API layer (FastAPI) and relational DB
b. Relationships and interactions
1. At startup, Streamlit app reads config.yaml via utils
2. Streamlit UI triggers pipeline scripts in src/ (or loads cached outputs from models/)
3. Data flows:
raw data → data_cleaning → interim/processed files → rfm_segmentation → segments output → basket_analysis → association_rules → recommendation → displayed charts
4. Visualization utilities produce Plotly graphs fed into Streamlit
5. Models folder acts as a persistent cache between pipeline runs
c. Architectural patterns & principles
• Modular separation: config, data-processing, UI, assets, model outputs
• Pipeline pattern: sequential ETL + ML modules
• Single-page interactive UI via Streamlit
• Clear separation of concerns: utils for logging/config, core logic in src, presentation in app/
d. Technologies / frameworks
• Python 3.x with pandas, numpy, scikit-learn, mlxtend, plotly
• Streamlit for frontend
• PyYAML for config
• File-based “database” (CSV / JSON), with future option for FastAPI + PostgreSQL