Analyzing device similarity

• US 9,460,390 B1
• Filed: 12/21/2011
• Issued: 10/04/2016
• Est. Priority Date: 12/21/2011
• Status: Active Grant

• Current Assignee: Emc IP Holding Company LLC (Dell Technologies Inc.)
• Original Assignee: EMC Corporation (Dell Technologies Inc.)
• Inventors: Lin, Derek, Kaufman, Alon, Villa, Yael
• Primary Examiner(s): Chaki, Kakali
• Assistant Examiner(s): Wu, Fuming
• Application Number:
  US13/332,889
• Time in Patent Office: 1,749 Days
• Field of Search: None
• US Class Current: 1/1
• CPC Class Codes: G06F 16/35 Clustering; Classification
• CPC Class Codes: G06N 20/00 Machine learning
• CPC Class Codes: G06N 5/04 Inference or reasoning models
• CPC Class Codes: G06N 7/01 Probabilistic graphical models, e.g. probabilistic networks
• CPC Class Codes: G06Q 10/08 Logistics, e.g. warehousing, loading or distribution; Inventory or stock management
• CPC Class Codes: G06Q 30/00 Commerce

Abstract

A method is used in analyzing device similarity. Data describing a device is received and a similarity analysis is applied to the data. Based on the similarity analysis, a measure of similarity between the device and a previously known device is determined.

Claims

1. A method for use in analyzing device similarity, the method comprising:

receiving data describing a set of devices, wherein the set of devices includes an unknown device and a previously known device, wherein the data includes a plurality of components associated with the set of devices, wherein the components include device hardware element data and application data, wherein each component of the plurality of components is measured by weight of popularity and frequency, and wherein the weight of each component of the plurality of components changes dynamically based on changing of the popularity and the frequency of use of the plurality of components;

based on the data, collecting unlabeled pairs of components in connection with the set of devices in which each pair is observed matching status for each component, wherein the said collecting enables preparation of multi-dimensional vectors for use in connection with training vectors stored in a matrix, wherein each component is represented in the multi-dimensional vectors by the group consisting of matching components, mismatching components, and missing components;

projecting, by a principal component analysis using singular value decomposition, the matrix to a lower dimensional space or latent space;

based on a cosine similarity angle, determining a pair of vectors maximally apart in the latent space;

determining, from the pair, the vector corresponding to a high dimensional vector where all or nearly all components match;

utilizing the said determined vector as an origin; and

determining a deviation from the origin for defining a device match similarity score.

2. The method of claim 1, wherein the measure is used to identify whether a user is accessing from known detected device.

3. The method of claim 1, wherein the measure is used for e-commerce.

4. The method of claim 1, wherein a data-driven modeling framework detects probabilistically whether the device is previously known device.

5. The method of claim 1, wherein depending on the measure of similarity, the device is classified as the same as previously known device.

6. The method of claim 1, wherein the measure of similarity is based on offline automatic training from web data.

7. The method of claim 1, wherein the measure of similarity accounts for importance of an element based on a frequency of the element in a population.

8. The method of claim 1, wherein the measure of similarity accommodates new device element additions.

9. The method of claim 1, wherein the measure of similarity is based on an unsupervised learning method.

10. A system for use in analyzing device similarity, the system comprising:

first logic receiving data describing a set of devices, wherein the set of devices includes an unknown device and a previously known device, wherein the data includes a plurality of components associated with the set of devices, wherein the components include device hardware element data and application data, wherein each component of the plurality of components is measured by weight of popularity and frequency, and wherein the weight of each component of the plurality of components changes dynamically based on changing of the popularity and the frequency of use of the plurality of components;

based on the data, second logic collecting unlabeled pairs of components in connection with the set of devices in which each pair is observed matching status for each component, wherein the said collecting enables preparation of multi-dimensional vectors for use in connection with training vectors stored in a matrix, wherein each component is represented in the multi-dimensional vectors by the group consisting of matching components, mismatching components, and missing components;

third logic projecting, by a principal component analysis using singular value decomposition, the matrix to a lower dimensional space or latent space;

based on a cosine similarity angle, fourth logic determining a pair of vectors maximally apart in the latent space;

fifth logic determining, from the pair, the vector corresponding to a high dimensional vector where all or nearly all components match;

sixth logic utilizing the said determined vector as an origin; and

seventh logic determining a deviation from the origin for defining a device match similarity score.

11. The system of claim 10, wherein the measure is used to identify whether a user is accessing from known detected device.

12. The system of claim 10, wherein the measure is used for e-commerce.

13. The system of claim 10, wherein a data-driven modeling framework detects probabilistically whether the device is previously known device.

14. The system of claim 10, wherein depending on the measure of similarity, the device is classified as the same as previously known device.

15. The system of claim 10, wherein the measure of similarity is based on offline automatic training from web data.

16. The system of claim 10, wherein the measure of similarity accounts for importance of an element based on a frequency of the element in a population.

17. The system of claim 10, wherein the measure of similarity accommodates new device element additions.

18. The system of claim 10, wherein the measure of similarity is based on an unsupervised learning method.