Data rows
—
Available after an original CSV is inspected.
Dataset-first Python AI/ML project
Explainable AI-Based Retail Customer Behavior Prediction and Personalized Recommendation System
A guarded project workspace for an executable Python pipeline: inspect the real retail schema first, then run only the analytics, prediction, recommendation, and explanation modules your dataset can truthfully support.
Columns
—
No schema has been assumed.
Missing cells
—
Calculated from the actual uploaded file.
Duplicate rows
—
Exact duplicate detection runs at inspection time.
Inspect your retail dataset
Upload the original CSV to inspect its real schema in this workspace. The accompanying Python pipeline accepts both CSV and Excel files when you run it locally.
Drop one CSV file here or choose it above
No customer or purchase metrics are assumed before inspection.
No dataset inspected yet
Start by selecting the original retail CSV. This project will map only columns it can actually find and will clearly flag any unavailable AI/ML module.
Executable pipeline readiness
Every module is dataset-gated
The local Python implementation saves genuine figures, metrics, tables, SHAP artifacts, and a final report only when its prerequisite columns and data sufficiency checks pass.
Awaiting dataset
Customer behavior features
Builds recency, frequency, monetary value, average order value, quantity, product diversity, and purchase cadence at customer level.
Requires customer, date, and transaction value signals.
Awaiting dataset
RFM segmentation
Selects an appropriate K-Means cluster count using elbow and silhouette diagnostics, then derives human-readable segment labels from cluster behavior.
Requires the customer behavior feature dataset.
Awaiting dataset
Time-based churn prediction
Creates a future-inactivity target from the transaction timeline and compares Logistic Regression, Random Forest, and XGBoost without data leakage.
Requires customer transactions over enough time.
Awaiting dataset
Product recommendation
Uses Apriori association rules for market baskets and, where interactions allow, an item-based collaborative filtering signal for recommendations.
Requires a product field plus a transaction ID or customer-date basket key.
Awaiting dataset
Explainable AI
Applies SHAP to the selected churn model and saves global importance, a SHAP summary, and a real individual-customer explanation.
Depends on a viable time-based churn model; no explanations are fabricated.
Local execution
Run the pure Python project
Use the included retail_xai_pipeline.py script with your original dataset. It supports CSV and Excel input, writes outputs to a folder you choose, and records skipped modules and limitations in the generated report.
Command
python retail_xai_pipeline.py --input your_retail_dataset.csv --output outputsModel safety: churn labels use a temporal split, and model comparison is skipped rather than faked when the data cannot form both classes.
Business output: the final customer intelligence table joins supported segments, churn risk, and recommendations without filling unsupported columns with fabricated values.