Methods, systems and computer code for finding outlier hyper cells in multidimensional hierachical data sets

1. A computer program product encoding a computer program stored on a non-transitory computer readable storage medium for executing a set of instructions by a computer system comprising one or more computer processors for carrying out a method for identifying outlier hyper-cells in an N-dimensional hierarchical dataset having dimensions {x 1 , . . . X N }, the method comprising:

a) given a plurality P GIVEN of hypercells {h 1 , . . . h m } for the N dimensions, each said hypercell h i P GIVEN having a respective participating-dimension set D i ⊂{x 1 , . . . X N }, for said each hypercell h i of said given plurality P GIVEN , categorizing, in accordance with a respective outlier score of said each hypercell hi, said each hypercell hi as one of:

i) a likely outlier hypercell; and

ii) a less-likely outlier hypercell, such that:

i) a first categorized plurality P 1 CAT of said hypercells is categorized as likely-outliers;

and ii) a second categorized plurality P 2 CAT of said hypercells is categorized as less-likely-outliers;

b) selecting a sub-plurality P SUB ⊂P 2 CAT of hypercells {k 1 . . . k L } from said second categorized plurality P 2 CAT of less-likely-outliers in accordance with said outlier scores of hypercells said sub-plurality P SUB ⊂P 2 CAT of hypercells;

c) from said sub-plurality P SUB of hypercells, forming a one-level-extension hypercell set of P 1 EXT including, for each hypercell k j of P SUB , a respective at least one one-level extension of said each hypercell k j of P SUB ;

d) defining a set P′ GIVEN which includes at least some hypercells of P 1 EXT such that there is at least one one-level-extension hypercell hc, which is a one-level extension of a member cell mc mc P 2 CAT that is included in P′ GIVEN only if

i) hc is not a previously considered hypercell;

ii) hc is not a subcell of a previously categorized likely-outlier hypercell; and

e) assigning P GIVEN =P′ GIVEN and repeating steps (a)-(d).

2. The computer program product of claim 1 wherein hypercells of P GIVEN are dimensionally inhomogenous.

3. The computer program product of claim 1 wherein hypercells of P GIVEN are dimensionally homogenous.

4. The computer program product of claim 1 wherein a majority of hypercells of P GIVEN are primitive hypercells where every dimension participates.

5. The computer program product of claim 1 wherein P GIVEN includes primitive hypercells from a majority of dimensions of the N dimensions of the dimensions {x 1 , . . . x N }.

6. The computer program product of claim 1 wherein P GIVEN includes a majority of possible primitive hypercells.

7. The computer program product of claim 1 wherein P GIVEN is a majority spanning plurality of hypercells.

8. The computer program product of claim 1 wherein said one-level extension of step (c) is dimensionally homogeneous.

9. The computer program product of claim 1 wherein said one-level extension of step (c) is dimensionally in homogeneous.

10. The computer program product of claim 1 wherein P′ GIVEN includes, for at least one hypercell of P 2 CAT, a majority of possible one-dimensional extensions which are not previously considered hypercells and not subcells of a previously categorized likely-outlier hypercell.

11. The computer program product of claim 1 wherein:

i) the hypercells are descriptive of discrepancies between two metering systems; and

ii) the method further comprises:

e) in accordance with identities of the categorized outlier hypercells of a plurality of iterations of step (b) providing a representation of the categorized outlier hypercells to a user.

12. The computer program product of claim 1 wherein:

i) the hypercells are descriptive of discrepancies between two metering systems; and

ii) the method further comprises:

e) in accordance with identities of the categorized hypercells of a plurality of iterations of step (b), billing or crediting a subscriber.

13. The computer program product of claim 1 wherein:

i) the hypercells are descriptive of discrepancies between two metering systems; and

ii) the method further comprises:

e) in accordance with identities of the categorized hypercells of a plurality of iterations of step (b), adjusting an operating parameter of at least one of the two metering systems.

14. The computer program product of claim 1 wherein both metering systems measures at least one of water consumption, oil consumption, gas consumption, electricity consumption and telecommunications services consumption.

15. A system for identifying outlier hyper-cells in an N-dimensional hierarchical dataset having dimensions {x 1 , . . . X N }, the system comprising:

a) a data storage for storing a representation of a plurality P GIVEN of hypercells {h 1 , . . . h m } for the N dimensions, each said hypercell h i P GIVEN having a respective participating-dimension set D i ⊂{x 1 , . . . X N } for said each hypercell hi of said given plurality P GIVEN ;

b) a hypercell categorizer for categorizing, in accordance with a respective outlier score of said each hypercell h i , said each hypercell h i as one of:

i) a likely outlier hypercell; and

ii) a less-likely outlier hypercell,

such that;

i) a first categorized plurality P 1 CAT of said hypercells is categorized as likely-outliers; and

ii) a second categorized plurality P 2 CAT of said hypercells is categorized as less-likely-outliers;

c) a hypercell selector operative to select a sub-plurality P SUB ⊂P 2 CAT of hypercells {k 1 , . . . k L } from said second categorized plurality P 2 CAT of less-likely-outliers in accordance with said outlier scores of hypercells said sub-plurality P SUB ⊂P 2 CAT of hypercells;

d) a hypercell extender operative to form from said sub-plurality P SUB of hypercells, a one-level-extension hypercell set of P 1 EXT including, for each hypercell k j of P SUB , a respective at least one one-level extension of said each hypercell k j of P SUB ;

e) a recursion set selector operative to define a set P′ GIVEN which includes at least some hypercells of P 1 EXT such that there is at least one one-level-extension hypercell hc, which is a one-level extension of a member cell mc mc P 2 CAT that is included in P′ GIVEN only if:

i) hc is not a previously considered hypercell;

ii) hc is not a subcell of a previously categorized likely-outlier hypercell; and

f) a recursion manager operative to:

A) pass the representation of P GIVEN to the hypercell categorizer and receive a representation of P 2 CAT from the hypercell categorizer;

B) pass the received representation of P 2 CAT to the hypercell selector and receive a representation of P SUB from the hypercell selector;

C) pass the received representation of P SUB to the hypercell extender and receive a representation of P 1 EXT from the hypercell extender

D) pass the representation of P 1 EXT to the recursion set selector and to receive a representation of P′ GIVEN from the recursion set selector 350 ;

E) set P GIVEN =P′ GIVEN in the memory and repeat steps (A)-(D).

16. The system of claim 15 where the hypercells are descriptive of discrepancies between two metering systems, and the system further comprises:

g) a user output module configured, in accordance with identities of the categorized outlier hypercells of a plurality of iterations of step (b), to provide a representation the categorized outlier hypercells to a user.

17. The system of claim 15 where the hypercells are descriptive of discrepancies between two metering systems, and the system further comprises:

g) a billing adjuster configured, in accordance with identities of the categorized outlier hypercells of a plurality of iterations of step (b), to bill or credit a subscriber.

18. The system of claim 15 where the hypercells are descriptive of discrepancies between two metering systems, and the system further comprises:

g) a meter adjuster configured, in accordance with identities of the categorized outlier hypercells of a plurality of iterations of step (b), to adjust an operating parameter of at least one of the metering systems.