Problem/Usefulness of Solution
- Credit card repayment continues to be a problem plaguing financial institutions. Ability to predict the repayment status of credit cards is essential to financial institution’s ability to manage its credit portfolio and grant new credit limits to customers, ultimately contributing to the profit of the financial institution. There is no singular factor that can correctly predict the repayment outcome of credit card. Several factors are often needed to make a plausible inference on the outcome of the repayment. Applying Machine Learning Algorithms on several variables, mostly from historical data and customer attributes, we can predict the outcome of credit card repayment with a higher probability. These models can be deployed on customer data to correctly determine credit outcomes with its attendant cost-saving benefits and profit.
Data
- The data comes from real banking dataset used in the Risklab hackathon organized by Risklab in 2018. The dataset used for analysis was a real customer banking data and have been anonymized to protect customer information. Dataset includes variables about customer attributes and historical credit information for each customer. The dataset was used for exploratory analysis of variable relationships and predictive analysis of repayment outcome using classification techniques.
Look at the "Reports" for detailed info regarding methodology and results.