Multi source Customer Mart for Female Recommendations in Marketplace (E-Commerce |Recommendations)

Description

This dataset represents a Gold Layer customer feature mart designed for female-oriented e-commerce recommendation systems. It includes 5,878 anonymized customer records with 17 engineered features derived from large-scale transactional, product, and review data. Built using a data warehouse pipeline (extraction, cleaning, integration, and transformation), the dataset provides a denormalized, machine-learning-ready representation of customer behavior, spending patterns, retention signals, and product preferences. It originates from a star-schema model integrating customer, orders, product, time, and geographic dimensions. Each record represents a unified customer profile supporting personalization, recommendation systems, churn prediction, customer segmentation, and customer lifetime value analysis. Key behavioral features include Recency, Frequency, and Total Orders. Monetary features include Total Spend, AvgOrderValue, and AvgPricePerItem. Preference features capture TotalItems, UniqueProducts, and quantity behavior. Retention features include CustomerLifespanDays and ChurnLabel. Profiling features include Country, Region, and JoinDate. All customer identities are anonymized, and sensitive attributes are removed. The dataset is designed for research and industrial applications in e-commerce analytics, with emphasis on female-focused recommendation systems and personalization.

Steps to reproduce

1. Data Acquisition: Raw retail transactions, product reviews, and customer-related data were collected from open-source sources and ingested into the Bronze Layer. 2. Data Cleaning and Standardization: Missing values, duplicates, noisy records, inconsistent formats, and invalid transactions were detected and corrected. Customer identifiers were anonymized and sensitive fields removed. 3. Silver Layer Integration: Cleaned sources were integrated and transformed into structured dimensional entities including customer, product, time, location, order, and review data. 4. Dimensional Modeling: A Star Schema was designed with dimension and fact tables to support analytical consistency and scalable querying. 5. Feature Engineering: Gold Layer features were generated, including RFM metrics (Recency, Frequency, Monetary), order statistics, diversity metrics, churn labels, lifespan indicators, pricing metrics, and regional attributes. 6. Customer Feature Mart Construction: Engineered features were consolidated into a denormalized customer-level flat dataset where each row represents one customer analytical profile. 7. Validation and Quality Checks: Feature distributions, business rules, consistency constraints, and derived metrics were validated to ensure reliability. 8. Gold Layer: Final curated features were exported as CSV and documented for machine learning, recommendation, and warehouse analytics use.

Institutions

• Mansoura University

  Dakahlia, Al Mansurah
• New Mansoura University

  Dakahlia, Al Mansurah

Categories

Related Links

Licence