Data Analysis Using Parent-Child Factors for Causal Ranking

The patent introduces an intermediary mechanism (parent-child relationship structure) between the data elements and the regression analysis. By organizing data elements into hierarchical parent-child relationships, the system mediates the selection process, allowing only parent elements to be used as explanatory variables. This intermediary structure filters out redundant child elements that would otherwise create multiple equivalent models, thus resolving the complexity of model selection while maintaining comprehensive factor coverage.

The patent segments the data elements into hierarchical levels (parent elements and child elements). By dividing the explanatory variables into these segments, the system can evaluate child elements (which may have many levels and high fitness) against their parent elements. This segmentation allows the identification of parent elements that represent broader categories, ensuring that important factors with fewer levels are not overlooked even when child elements with many levels show better model fitness.

Instead of holding down the scores of data elements with many levels (conventional approach), the patent inverts the approach by using parent-child relationships to adjust rankings. The system calculates the difference in fitness between child elements and their parent elements, and uses this inverted logic to determine final rankings. This ensures that even if child elements naturally have better fitness due to having more levels, their parent elements can be identified as the true factors, correcting the ranking to reflect actual importance rather than just model fitness.