Method for capturing local and evolving clusters
View Patent ↗The present invention provides a method of detecting the growth and development of clusters in a data set. The data set is divided into a number of slices and an algorithm is applied to the data held in each data slice set. Each slice can be compared with the subsequent slice to determine which clusters persist from slice to slice. Random data agglomerations in a single slice may give the appearance of a cluster but their random nature means that they are unlikely to persist so those clusters that persist across a number of slices, or that show the strongest measure of persistence, are most likely to represent a data cluster that represents a situation of interest.
1. A method of evaluating one or more clusters in a data set, the method comprising:
using at least one computer having a processor coupled to a program memory containing a computer program configured to:
a) divide said data set into a plurality of data slices;
b) select a plurality of data slice groups from said plurality of data slices;
c) apply a data clustering algorithm to each of said data slice groups;
d) for each of said data slice groups, compare the results of the data clustering algorithm with the results obtained from a subsequent data slice group to determine one or more persistent data clusters; and
e) store in a memory device said one or more persistent data clusters determined in d).
2. A method according to claim 1 , wherein in d) a data cluster is defined as a persistent data cluster if it is present in a plurality of subsequent data slice groups.
3. A method according to claim 2 , wherein in d) a data cluster is defined as a persistent data cluster if it is present in at least 3 subsequent data slice groups.
4. A method according to claim 1 , wherein in d) a data cluster is defined as a persistent data cluster if its cluster persistence parameter exceeds a given threshold.
5. A method according to claim 4 , wherein the cluster persistence parameter comprises a weighted average of a number of values.
6. A method according to claim 4 , wherein said cluster persistence parameter represents the proportion of cluster data present in a cluster in one data slice group and in a subsequent data slice group.
7. A method according to claim 6 , wherein the cluster persistence parameter comprises a weighted average of a number of values.
8. A non-transitory computer readable medium tangibly storing computer executable code which when loaded onto a computer will execute the method of claim 1 .