Customer-Behaviour-Prediction

Customer Segmentation for Marketing, RFM Analysis, Purchase Day Prediction

Problem Statement:

We help Rainbow Store understand and predict customer behaviours. Here we are dealing with store’s basket-level transaction data. The store manager wants to know which customers are likely to purchase next month so he can prepare his marketing campaign.

1. Implement a model that predicts which customers make at least 1 purchase in a given month using features generated from the 2 previous months. For example, data from February and March can be used to predict purchases in April; data from March and April can be used to predict purchases in May.

2. You're asked which customers to send promotional e-mails to next month, based on your model. What is your recommendation?

Data:

• csn, date, salesqty, price/unit, article_id
• We have total of 5 months of transaction data

Feature Engineering:

• RFM
• Customer Segments Calculated using Cluster Analysis
• Last 5 purchase days per customer
• Distribution (mean, std) of purchase day difference
• Target label: Purchase day in given month

Modelling:

• XGBoost (Final Model)
• K-fold Cross Validation

Evaluation:

• Accuracy
• Train: 83%, Test: 78%