sarveshchhetri/E-commerce_return_abuse_detection_dataset

In the booming world of e-commerce, hassle-free return policies are a major driver of customer satisfaction. However, this convenience comes at a massive cost: Return Abuse and Fraud. Retailers lose billions annually to "wardrobing" (wearing an item once and returning it), policy abuse (excessive legitimate returns), and outright fraudulent returns (returning empty boxes, stolen goods, or different items).

Detecting these bad actors without adding friction to the experience of legitimate customers is one of the most critical challenges in modern retail data science.

This dataset is a synthetic dataset that contains 60,000 highly detailed e-commerce return transactions, meticulously featuring customer demographics, behavioral metrics, and shipping information. The target variable is abuse_type, an imbalanced multi-class label categorized into:

• Legitimate (70.1%): Normal, honest returns.
• Policy Abuser (11.9%): Customers who exploit return policies (e.g., returning 90% of what they buy).
• Fraudulent Return (10.2%): Malicious returns (e.g., sending back empty boxes, tracking fraud).
• Wardrobing (7.7%): Buying items for a one-time event and returning them.

With zero missing values and a rich set of 35 features, this dataset is an exceptional playground for feature engineering, anomaly detection, and predictive modeling.

🎯 Starter Task Ideas (How to use this dataset)

1. Multi-Class Classification: Build an XGBoost or LightGBM model to predict the exact abuse_type (or abuse_label). Challenge yourself to handle the class imbalance using techniques like SMOTE or class-weighting!
2. Fraud Anomaly Detection: Treat the dataset as an unsupervised learning problem. Can algorithms like Isolation Forests or Autoencoders identify the 10.2% of Fraudulent Returns?
3. Exploratory Data Analysis (EDA): Does "wardrobing" happen more often in specific product_category groups (like clothing)? Do policy abusers have higher account_age_days than legitimate customers?
4. Time-Series Analysis: Calculate the time deltas between order_date and return_date to find seasonality or patterns in malicious returns.

---