Complexity-Based Propensity Score Model Selection With Stop Criteria

The patent segments the evaluation process by dividing propensity score models into different complexity groups (e.g., simple, moderate, complex). The system evaluates models in sequential batches according to their complexity rather than evaluating all models simultaneously. This segmentation allows the system to identify suitable models at earlier, less resource-intensive stages without compromising the comprehensiveness of model selection.

The patent applies partial action by evaluating only a subset of available propensity score models rather than all models. The system uses stop criteria to determine when sufficient models have been evaluated, allowing it to terminate the evaluation process early. This partial evaluation approach maintains reliable model selection while significantly reducing computer resource consumption compared to exhaustive evaluation of all models.

The patent applies preliminary action by pre-categorizing propensity score models into complexity groups before the evaluation process begins. Models are sorted and organized by complexity level in advance, allowing the sequential evaluation to proceed efficiently without requiring complex real-time decisions about which models to evaluate next. This preliminary organization reduces the overhead during the actual evaluation phase.

The patent changes the evaluation parameter from evaluating all models to evaluating models based on complexity thresholds. The system adjusts evaluation parameters dynamically, stopping when predefined criteria are met (such as finding a sufficient number of suitable models or reaching a maximum complexity level). This parameter-based approach optimizes the balance between resource consumption and evaluation time.