DATA PROCESSING TO ADDRESS UNREPRESENTATIVENESS

摘要

A method of processing data in a fixed taxonomy in order to address quantitative contradiction and/or unrepresentativeness in the data comprises, commencing at end points of the taxonomy: (i) for each set of parent and children nodes, determining a measure of the parent node properties by suitable (i.e. linear, or non-linear or other logical) combination of the properties of the children nodes and comparing this with the parent node properties in the taxonomy to locate quantitative contradictions between the properties of the parent node and children nodes; (ii) identifying a set of knowable data, the knowable data being a suitable representation of the data in a parent and children node set that is not currently known and that would rectify a perceived contradiction between the properties of the parent node and the children nodes found by comparison step (i); (iii) evaluating a balancing correction required for the knowable parent and children nodes in order to reduce the difference between the properties of the parent node and children nodes towards zero (or optionally to zero, if possible) and applying this correction to all sets of unknown but knowable parent and children node sets; (iv) moving up the taxonomy one level at a time, repeating steps (i) to (iii) taking account of the prior balancing corrections of the nodes subject to balancing corrections in step (iii) for the preceding level when making balancing corrections for ancestors of those nodes; and (v) once the most senior knowable parent node has been balanced, repeating steps (i) to (iii) commencing with the parent and children node set of the most senior knowable parent node and moving down the taxonomy one level at a time, wherein in step (iii) the children nodes are balanced with the parent node without any adjustment to the parent node properties, which remain fixed.