Deep learning survival analysis for consumer credit risk modelling - Jiahang Zhong, PhD

Jiahang Zhong, PhD was speaking at ODSC Europe 2019
→ To watch more videos like this, visit aiplus.odsc.com ←
In consumer lending and credit card business, accurate prediction of credit risk at individual level is critical for providing better products to the customers and establishing competitive advantages in the market. Traditional credit risk modelling has been focusing on the probabilistic prediction of outcomes, such as default or early settlement, over a particular time horizon. However, further prediction of the time of such events can lead to more precise risk estimate. Survival analysis has been widely adopted for decades across various industries for such so-called time-to-event prediction. The recent advancements in big data and deep learning further enabled more sophisticated survival model to achieve precise time-to-event predictions at individual level.
In this talk we are going to briefly review the theoretical concepts of survival analysis and a few classic models. We will then describe our approach to combine it with deep learning models, in the context of consumer lending.
Do You Like This Video? Share Your Thoughts in Comments Below
Also, You can visit our website and choose the nearest ODSC Event to attend and experience all our Trainings and Workshops:
odsc.com/europe/
odsc.com/california/
Sign up for the newsletter to stay up to date with the latest trends in data science: opendatascience.com/newsletter/
Follow Us Online!
• Facebook: / opendatasci
• Instagram: / odsc
• Blog: opendatascience.com/
• Linkedin: / open-data-science
• Learning Videos: learnai.odsc.com
#DeepLearning #ODSC #DataScience

[@danielbaena4691]

Do you know maybe if Jiahang has a paper associated with this talk? Thank you

[@zhaobryan4441]

could you plz share the slides

[@rechellejacobs3315]

The early repayment regarded as a competing risk is questionable, can you clarify why it is regarded as a competing risk, since this type of early settlement event will not cause the event of interest (which is default) to happen

[@mikemccabe3585]

Any reason for a customer to exit the portfolio is a risk, e.g. bad debt, early pay-off, refinance, etc. In regards to a loan portfolio, the segment of customers that pay-off early typically have a lower risk for default, e.g. they have the funds or can re-finance for a lower rate. The expected profitability of the portfolio is lowered as the portfolio is now weighted towards a population with higher default risk.