Purged K-Fold Cross-Validation - Quant Researcher Interview Question
Difficulty: Hard
Category: machine_learning
Asked at: Numerai, Citadel, WorldQuant, Two Sigma, Citadel Securities, Optiver, AQR Capital Management, Man Group
Topics: cross_validation, time_series, finance, data_processing
Problem Description
Standard K-Fold cross-validation is unsuitable for financial time series due to label overlap and serial correlation, which cause information leakage and optimistic performance estimates. Purged K-Fold Cross-Validation addresses these issues by removing training observations that overlap with the test set's label interval or immediately follow it. This method ensures that the model is trained on data strictly independent of the testing period, mimicking a realistic trading scenario.
Task
Implem
Practice this hard researcher interview question on MyntBit - the LeetCode for quants with 200+ quant interview questions for Jane Street, Citadel, Two Sigma, and other top quantitative finance firms.