Abstract:
|
Score developed by using statistical classification approaches based on individual profile characteristics is a basic tool for portfolio decision. Changes of economy, culture, or age, etc. would cause changes to profile characteristics for individuals or business entities, i.e., population shift. Population shift would, in turn, mean changes to decision-supporting algothms derived based on profile characteristics. Therefore, decisions based on scores, or other algorithms derived from profile chracteristics, are likely to be affected as any change occurs to individual profiles in a portfolio.
In this study, the impact of population shift on score-based decisions, including binary and proressive actions, is discussed by using simulated data. Also, sensitivity of single and fusion scores is discussed in relation to score-based decision. Finally, simple but practical diagnostics for population shift, and remedies for population shift, are suggested in practitioners' point of view.
|