Recommendation Engine Accounting for Preference Drift

Aspects of the disclosure relate to content prediction. A computing platform may train a collaborative recommendation engine to output recommendation information based on historical preference information and corresponding data drift. The computing platform may receive an account access request from a user device. The computing platform may identify, at a first time and using the collaborative recommendation engine, a preference group for the user device. The computing platform may receive, at a second time later than the first time, a second account access request from the user device. The computing platform may identify, using the collaborative recommendation engine, the preference group and data drift corresponding to the preference group between the first time and the second time, which may indicate a second set of preferences at the second time. The computing platform may generate, based on the second set of preferences, recommendation information for the user device.