How would you handle noisy market data when building predictive models for financial transaction or risk use cases?
