How tejasshinkar12/ecommerce-transactions-analysis-and-customer-segmentation works
Project Type and Purpose
• This is a data-analytics & visualization project built in Python, centered on:
• Ingesting raw CSV data (transactions, customers, products)
• Performing EDA, feature engineering, clustering, lookalike modeling
• Exposing results via a Streamlit dashboard
• Key goal: generate business-actionable insights and customer recommendations.
File Structure
• Top level:
• datasets/ – raw CSV inputs + derived customer_data.csv
• notebooks/ – Jupyter notebooks for EDA, Clustering, Lookalike
• app/ – Streamlit dashboard (streamlit_main.py + requirements.txt)
• Reports/ – exported artifacts (Lookalike.csv)
• .devcontainer/ – container config for reproducible environment
• Patterns: clear separation of data layer (datasets), processing/modeling layer (notebooks), presentation layer (app).
Extract Architecture Insights from README
• Data ingestion → cleaning → feature engineering yields an intermediate “customer_data.csv.”
• Clustering notebook builds RFM + PCA + K-means pipeline.
• Lookalike notebook computes cosine similarities → exports recommendations CSV.
• Streamlit app imports all artifacts, renders interactive visualizations (radar, treemap, 3D scatter) and tables.
Main Components & Their Interactions
a. Data Layer
• Raw Data Files (datasets/Customers.csv, Products.csv, Transactions.csv)
• Derived Data (datasets/customer_data.csv)
b. Processing Layer (Notebooks)
• EDA Notebook: merges raw files → cleans → feature engineering → writes customer_data.csv
• Clustering Notebook: reads customer_data.csv → PCA + K-means → outputs cluster assignments (in memory or as file)
• Lookalike Notebook: reads customer_data.csv (and clusters if needed) → cosine similarity → writes Reports/Lookalike.csv
c. Presentation Layer
• Streamlit App (app/streamlit_main.py)
• Reads datasets/customer_data.csv, cluster outputs, lookalike CSV
• Generates interactive visualizations with Plotly & Streamlit UI
d. Environment & Deployment
• .devcontainer/devcontainer.json for VSCode containerization
• requirements.txt listing Python dependencies
Architectural Patterns & Principles
• Modular pipeline: each notebook is a self-contained data-processing stage
• Clear separation: data → processing → modeling → presentation
• File-based artifacts as integration points between stages
• Reproducible dev environment via devcontainer
Essential Details to Capture
• File paths/folders as component names
• Data flows between notebooks and artifacts
• Tech stack per component (Python 3.x, Pandas, Scikit-learn, Plotly, Streamlit)
• Reproducibility layer (.devcontainer)
• Output artifact storage (Reports/)