| 摘要 |
A method and system for identifying outlier hyper-cells in anN-dimensional hierarchical datset having dimensions {x1, xN} are provided. In some embodiments, the method is applied recursively where (i) in accordance with outlier "scores," hypercells of a hypcell set PGIVEN are categorized S201 as "likely-outlier" or "winners" while other hypercells are (at least initially) rejected as a set of "losers"; (ii) scores of hypercells of a subset of the initially rejected "losers" are analyzed, and in accordance with these scores a subset of "winners of losers" are selected S209; (iii) one-level extensions S213 of at least some hypercells of these "winners of losers" are formed; (iv) a recussion set S226 is formed where at least one "extension" hypercell hc is included in the recusion set only if the hypercell is not a previously considered hypercell and/or a subcell of a previously categorized (and/or reported) as 'outlier' hypercell; (v) the aforementioned steps are repeated after setting PGIVEN to the recursion set. In some embodiments, the hypercells are indicative of discrepancies between measurements provided by two different metering systems. |
