Data Source: https://www.kaggle.com/shelvigarg/credit-card-buyers
Problem Statement: Need to drive Credit Card Acquisition cost efficiently by maximizing internal resources.
Objective: Predict which of existing eligble customer would be interested in applying for the credit card based on their demo, relationship with bank and exisiting product holdings.
Methodology:
- Graph will be used to plot and understand the relationship between the product interest and each variables. The output can be used to influence Marketing decision on segmentation and messaging for external customer acquisition.
- Supervised learning technique will be used (Logistic Regression / Random Forest Tree). The output can be used to cross-sell via telemarketing / EDM / SMS / Digital Advertising.
Outcome A propensity list with high accuracy for direct marketing - can I draw the graph
Flow:
What are the features
EDA - Analysis - what are the important features / relationship
Try more model XGBoost, svm, compare 4 models - use accuracy - True negative, etc.
potential if have time: Combine model
Evaluation on 4 models
Conclusion & Application (plot propensity graph)
Feature importance - rating which one has higher / lower - what are potentially other features that should be included
What are the insights (supplement)