1

Topic: Cluster analysis

All kind day. Probably not absolutely correctly selected section, but the given task I implement Oracle means. And so it is had the normalized fact table, on its basis splitting into clusters into the base k-means algorithm is led. It is necessary to implement sampling of the closest facts in a cluster at appearance of new record. I.e. having the new fact (or the hypothesis) to define gets to what cluster record and to what the close in a cluster to records can correlate. In advance thanks.

2

Re: Cluster analysis

borka1985 wrote:

All kind day. Probably not absolutely correctly selected section, but the given task I implement Oracle means. And so it is had the normalized fact table, on its basis splitting into clusters into the base k-means algorithm is led. It is necessary to implement sampling of the closest facts in a cluster at appearance of new record. I.e. having the new fact (or the hypothesis) to define gets to what cluster record and to what the close in a cluster to records can correlate. In advance thanks.

As a variant here the ready decision
https://docs.oracle.com/cd/E11882_01/da … m#DMCON562

3

Re: Cluster analysis

Alexander Ryndin, thanks! But the problem that the given mechanism finds a centroid and finds distance in each cluster to a centroid. It turns out that it is necessary to install an accessory to a cluster, and then to define what measurements get to it and to construct on restricted sampling the table of distances between a hypothesis.

4

Re: Cluster analysis

borka1985;

5

Re: Cluster analysis

dbms_photoshop, I thank! By search found your post, but while we on 11. The question most likely on a correctness of the selected algorithm, probably someone faced the decision of similar tasks, I will be glad to any council.