Reference no: EM132556935

Propensity to Lapse - Model building Assessment

Part 1. Concept

One of our clients has shared their raw customer transaction data with us with the intent to analyze and describe customer behavior and come up with fresh insights. The client runs a reward system where the customers collect points from flights or purchases from affiliated partners (e.g. department stores, electronics stores, etc.) which they can then redeem to book flights, hotels, or rent a car.

Here, after conducting the data and feature engineering process we have resulted in the dataset which you will find inside the Case Study folder. The State field (Label) is each customer's current lapsing status (Active or Lapsed converted to Boolean form - Active= 0, Lapsed=1).

1.1 Lapse Definition
When a customer does not have any collections or redemptions for a consecutive 12 months e.g. if a customer last collected in April 2015 and then didn't collect/redeem until May 2016, then they are categorized as lapsers.

1.2 Collections Definition
A customer interacts with the business partner to add loyalty points to their collection e.g. the customer purchases an item from a shop (which is a loyalty partner) and they win points that are added to their Balance.

1.3 Redemptions Definition
A customer uses some of the loyalty points they have in their Balance, e.g. they redeem X points to book a flight. The points redeemed are subtracted from the Balance.

Dataset
The given dataset consists of 5000 observations (customers) having the below fields (characteristics):
State: Lapsed status (Active= 0, Lapsed=1)
Sum_collect: how many times a customer collected
Sum_redeem: how many times a customer redeemed
Sum_collect_points: how many points a customer has collected in total
Sum_redeem_points: how many points a customer has redeemed in total
Years_in_the_program: years since customer's registration to the program
Months_since_last_transaction: months passed since customer's last action (collection or redemption)

Task
3.1 Build a Machine Learning Model, using the Model_Building_Guide script as a guide, successfully predicting if a customer with the given characteristics will Lapse in the next 12 months. You should write your code in Python.
• The model should be built in Python.
• If you feel like creating extra features from the provided dataset, feel free to do so.
• If you perform any data pre-processing, please document accordingly.
3.2 Create a PowerPoint presentation presenting your model and the key findings.
• The presentation should be aimed at business people with limited technical knowledge.
• Presentation should include:
o Overview of your methodology
o Findings / Insights - basically anything that will help us better understand churn.
o Recommendations on how to exploit this newly created model to gain business value
3.3. What we expect from you:
• Your model code
• Your PowerPoint presentation

Attachment:- Model building Assessment.zip