Label shift detection and adjustment in predictive modeling

One or more storage media storing instructions which, when executed by one or more processors, cause:

• for each scoring instance in multiple scoring instances in a scoring data set, using a first machine-learned model to generate a predicted label for said each scoring instance, wherein the first machine-learned model is trained using one or more machine learning techniques based on a plurality of training instances, each of which includes an observed label;
• generating a forecasting model based on time-series data that is based on first observed label data that corresponds to a first period of time;
• using the forecasting model to generate a forecast;
• performing a comparison between the forecast and second observed data that corresponds to a second period of time that is subsequent to the first period of time;
• detecting a shift in observed labels based on the comparison;
• in response to detecting the shift in observed labels, for each segment of one or more segments in a plurality of segments:
• identifying a portion of training data that corresponds to said each segment;
• for each training instance in a subset of the portion of training data:
• adjusting said each training instance to generate an adjusted training instance;
• adding the adjusted training instance to a final set of training data;
• using the one or more machine learning techniques to train a second machine-learned model based on the final set of training data.