I need python data science expert
Budget: $10 – $50 USD
Risk Segmentation for Car Loans Portfolio
Introduction
Customer segmentation involves categorizing the portfolio by industry, location, revenue, account size, and number of employees and many other variables to reveal where risk and opportunity live within the portfolio. Those patterns can then provide key measurable data points for more predictive credit risk management. Taking a portfolio approach to risk management gives credit professionals a better fix on the accounts, in order to develop strategies for better serving segments that present the best opportunities. Not only that, you can work to maximize performance
in all customer segments, even seemingly risky segments.
Customer segmentation analysis can lead to several tangible improvements in credit risk management: stronger credit policies, and improved internal communication and cooperation across teams.
Task scope
You are working in the retail risk modeling team and you are asked to build a risk-based segmentation model for car loans’ customers based on given historical data of customer behavior. The segmentation has to be from risk management perspective.
• Please watch this two videos with further explanation: https://www.youtube.com/watch?v=PLsUfDDytaE
• Risk-based segmentation: https://www.youtube.com/watch?v=2ZpLgUcucfQ
Project goals
Create a Python class using a DataFrame and several methods to automate the Risk-based segmentation process:
• Identify different potential segments sharing data (based on sharing missing values)
• Identify different segments of customers with different level of risk (the one explained in the second video)
• Research and recommend new ideas to risk segmentation
Tools
You are allowed to use Python only
Limitations
You are allowed to search for whichever information you need in the internet including but not limited to:
• Code syntax
• Business term (However you can ask me )
Please note that using an open source project which build the same required model considered as cheating.
Deliverables
• Project code/file that you used for your analysis and model building
Evaluation criteria
• Problem structuring
o How did you structure the problem?
o What assumptions did you make?
o How did you narrow the scope?
• Technical Skills
o How reliable (does it use your own class? does is it apply data quality controls?)), readable and flexible (can you apply your code to a new dataset?) was the code that you developed?
• Analytical Skills
o How logically sound, complete and meaningful was the approach (machine learning, statistics, analytics, visualization…) that you applied?
• Usefulness
o How useful would the results of your work for new datasets?
Data description
We will provide you with historical data of car loans. The data contains monthly status for each loan for 3 years. In addition to some demographic information
Notes:
• This data is Loan level NOT Customer level, meaning that one customer can take more than one loan
• The data is monthly starting from 2016-01-01 to 2019-09-01 so if the loan already started before Jan2016 you will find partial history for it.
• We have multiple programs under the car loans product
• Make sure you understand the difference between Buckets
*Data dictionary is provided in a separate file
Introduction
Customer segmentation involves categorizing the portfolio by industry, location, revenue, account size, and number of employees and many other variables to reveal where risk and opportunity live within the portfolio. Those patterns can then provide key measurable data points for more predictive credit risk management. Taking a portfolio approach to risk management gives credit professionals a better fix on the accounts, in order to develop strategies for better serving segments that present the best opportunities. Not only that, you can work to maximize performance
in all customer segments, even seemingly risky segments.
Customer segmentation analysis can lead to several tangible improvements in credit risk management: stronger credit policies, and improved internal communication and cooperation across teams.
Task scope
You are working in the retail risk modeling team and you are asked to build a risk-based segmentation model for car loans’ customers based on given historical data of customer behavior. The segmentation has to be from risk management perspective.
• Please watch this two videos with further explanation: https://www.youtube.com/watch?v=PLsUfDDytaE
• Risk-based segmentation: https://www.youtube.com/watch?v=2ZpLgUcucfQ
Project goals
Create a Python class using a DataFrame and several methods to automate the Risk-based segmentation process:
• Identify different potential segments sharing data (based on sharing missing values)
• Identify different segments of customers with different level of risk (the one explained in the second video)
• Research and recommend new ideas to risk segmentation
Tools
You are allowed to use Python only
Limitations
You are allowed to search for whichever information you need in the internet including but not limited to:
• Code syntax
• Business term (However you can ask me )
Please note that using an open source project which build the same required model considered as cheating.
Deliverables
• Project code/file that you used for your analysis and model building
Evaluation criteria
• Problem structuring
o How did you structure the problem?
o What assumptions did you make?
o How did you narrow the scope?
• Technical Skills
o How reliable (does it use your own class? does is it apply data quality controls?)), readable and flexible (can you apply your code to a new dataset?) was the code that you developed?
• Analytical Skills
o How logically sound, complete and meaningful was the approach (machine learning, statistics, analytics, visualization…) that you applied?
• Usefulness
o How useful would the results of your work for new datasets?
Data description
We will provide you with historical data of car loans. The data contains monthly status for each loan for 3 years. In addition to some demographic information
Notes:
• This data is Loan level NOT Customer level, meaning that one customer can take more than one loan
• The data is monthly starting from 2016-01-01 to 2019-09-01 so if the loan already started before Jan2016 you will find partial history for it.
• We have multiple programs under the car loans product
• Make sure you understand the difference between Buckets
*Data dictionary is provided in a separate file