K-means clustering using t-test computation
AssigneeIntel Corporation
InventorQian Diao
Life of this patent
Estimating the term…
Abstract
A method, apparatus, and system are provided for k-means clustering using t-test computation. According to one embodiment, k-means clustering is performed on a dataset. In performing k-means clustering, the dataset having various points is partitioned into several clusters. The closeness of a given point to a given cluster is determined. Then, t-test computation is performed to determine statistical linear relationship between the point and the cluster. If the point is found to be close to the cluster and a statistical linear relationship is found between the point and the cluster, the current position of the point is maintained.
Claims22
A method, comprising:
- performing k-means clustering of a dataset having a plurality of points, wherein the performing of the k-means clustering includes partitioning of the dataset into a plurality of clusters;
- determining closeness of a point of the plurality points to a cluster of the plurality of clusters;
- performing t-test computation to determine statistical linear relationship between the point and the cluster;
- and
- maintaining a current position of the point in the dataset if the point is close to the cluster, and if the point has a statistical linear relationship with the cluster.
An apparatus, comprising:
- a storage medium to store sets of instructions to perform k-means clustering using t-test computation;
- and
- a processor coupled with the storage medium, the processor to perform the k-means clustering of a dataset having a plurality of points,
- wherein the performing of the k-means clustering includes partitioning of the dataset into a plurality of clusters,
- determine closeness of a point of the plurality points to a cluster of the plurality of clusters,
- perform the t-test computation to determine statistical linear relationship between the point and the cluster, and
- maintain a current position of the point in the dataset if the point is close to the cluster, and if the point has a statistical linear relationship with the cluster.
A system, comprising:
- a k-means module to perform k-means clustering of a dataset having a plurality of points, wherein the performing of the k-means clustering includes partitioning of the dataset into a plurality of clusters;
- a Pearson Correlation Coefficient module to determine closeness of a point of the plurality points to a cluster of the plurality of clusters;
- a t-test module to perform t-test computation to determine statistical linear relationship between the point and the cluster;
- and
- a positioning module to maintain a current position of the point in the dataset if the point is close to the cluster, and if the point has a statistical linear relationship with the cluster.
A machine-readable medium having stored thereon data representing sets of instructions which, when executed by a machine, cause the machine to:
- perform k-means clustering of a dataset having a plurality of points, wherein the performing of the k-means clustering includes partitioning of the dataset into a plurality of clusters;
- determine closeness of a point of the plurality points to a cluster of the plurality of clusters;
- perform t-test computation to determine statistical linear relationship between the point and the cluster;
- and
- maintain a current position of the point in the dataset if the point is close to the cluster, and if the point has a statistical linear relationship with the cluster.
Dependencies are inferred from the claim text.