Contact Center Assignment Validation Using Weighted Historical Outcomes

The patent segments the historical data into distinct training set and validation set portions. The training set is used to train candidate assignment strategy models, while the validation set is used to validate them. This segmentation prevents overfitting by ensuring that models are tested on data they have not seen during training, thereby enabling accurate detection of overfitting and improving validation result accuracy.

The patent applies parameter changes by using different weighting schemes for training and validation. Specifically, it uses a first weighting scheme for training the models and a second weighting scheme for validating them. This allows the system to maximize the utility of limited representative historical data by optimizing how it is distributed and weighted across training and validation processes, thereby improving validation accuracy without requiring additional data.

The patent applies preliminary action by pre-establishing the validation set and weighting schemes before model training begins. The validation set is prepared in advance from historical data, and the weighting schemes are defined beforehand. This preliminary preparation streamlines the subsequent model training and validation process, allowing multiple candidate models to be evaluated efficiently without excessive computational complexity, thereby maintaining high model selection accuracy.