Skip to content

Identify which customer is willing to possess the insurance policy, so we campaign efficiently.

License

Notifications You must be signed in to change notification settings

nvlinhvn/customer-targeting-for-insurance

Folders and files

NameName
Last commit message
Last commit date

Latest commit

 

History

7 Commits
 
 
 
 
 
 
 
 
 
 
 
 
 
 

Repository files navigation

Customer Targeting for Insurance Promo

Identify which customer (~800-1000 users) is willing to possess the insurance policy, so we campaign efficiently.

Data and Privacy

  • Due to data privacy, all of attributes (85 attributes) and ID are completely anonymized
  • Attribute 1-42: Sociodemographic (All customers living in areas with the same zip code have the same socio-demographic attribute)
  • Attribute 43-85: Product Ownership

Problem Formulation

Imnbalance binary classification. The expected output is the probability of a user (or score) who is willing to buy the insurance

Feature Overview

  • Positive Correlation:
    • Customer sybtype and main type (1 & 5)
    • Household size and househould with children
    • Product ownership: Contribution to a product and corresponding amount (43-85)
  • Negative Correlation:
    • Rented House vs Home owners (30 & 31)
    • National Health Services and Private Health Insurance (35 & 36)
  • Potentially some features are mutually exclusive/inclusive

Feature Selection

Select a useful subset of features for modeling: 47, 59, 68, 1, 5, 42, 43, 37, 18, 44

Stacking Modeling

Metrics

  • We see users who tends to purchase insurance are in upper-middle class with higher income, contributing significantly, and possesing high number of car/fire policies