Data-drift emerged as a major challenge in 2020 for companies using predictive models to forecast consumer behaviour.
Almost a year after the pandemic began, many companies are still struggling to pin down consumer behaviour accurately. The disruption caused by the pandemic skewed all consumer predictive models and forced businesses to update their machine learning models regularly.
A Harvard Business Review report suggests that looking at past economic shocks can help companies fix unstable predictable models. Organisations can also embrace ensemble modelling that combines predictions from different models to present a more practical and reasonable range.
Capgemini’s Dan Simion said companies need to develop new machine learning models, which are nimble enough to increase accuracy by using new data on evolving consumer behaviour. Simion added, “Until there is some sort of stability, it will continue to be difficult for organisations to identify consistent trends.”
[4 minute read]