Data Distribution Shift
The phenomenon in which the statistical distribution of training data differs from that of production data. In finance it is especially pronounced because of data de-identification, sampling and synthesis, and it requires monitoring and rollback mechanisms.
Timeline (last 90 days)
Because financial data is highly sensitive, developers can often only use desensitized, sampled, or synthetic data, and the joint distribution of such data and production data may differ systematically.
Data distribution shift may cause zero-training prediction to underperform models trained specifically for the task, especially when business data has strong domain characteristics (such as medical coding or rare events in financial risk control).
All Facts (2)
Because financial data is highly sensitive, developers can often only use desensitized, sampled, or synthetic data, and the joint distribution of such data and production data may differ systematically.
50%UnverifiedData distribution shift may cause zero-training prediction to underperform models trained specifically for the task, especially when business data has strong domain characteristics (such as medical coding or rare events in financial risk control).
50%