Ranking data analytics results using composite validation

Patent metadata

Field	Value
Publication number	US-10210461-B2
Application number	US-201414222143-A
Country	US
Kind code	B2
Filing date	Mar 21, 2014
Priority date	Mar 21, 2014
Publication date	Feb 19, 2019
Grant date	Feb 19, 2019

Abstract

A method for performing assisted knowledge discovery includes receiving a dataset. Each of a plurality of analytical techniques is applied to the received data set to generate a plurality of corresponding analytical results. A composite validation metric is applied to each of the plurality of analytical results. The composite validation metric is a single scoring/ranking function that is created from a plurality of different scoring/ranking functions. The plurality of analytical results is presented to a user arranged in accordance with the results of the applying the composite validation metric to each of the plurality of analytical results. A selection from the user from among the plurality of analytical results is recorded. The user's selection is used to modify the composite validation metric such that the analytical techniques responsible for generating the selected analytical result is scored/ranked more highly.

First claim

What is claimed is:

1. A method for performing assisted knowledge discovery, comprising:
receiving a dataset;
applying each of a plurality of different analytical techniques to the received data set to generate a plurality of corresponding analytical results;
receiving a composite validation metric that is a single scoring or ranking function that is created from a plurality of different scoring or ranking functions, each of which is assigned a weighing that determines its relative influence on the composite validation metric;
modifying the received composite validation metric by removing therefrom each of the plurality of different scoring or ranking functions that is assigned a weighing that is less than a predetermined value;
applying the modified composite valuation metric to each of the plurality of analytical results;
displaying the plurality of analytical results to a user arranged in accordance with the results of the applying the modified composite validation metric to each of the plurality of analytical results;
recording a selection from the user from among the plurality of analytical results; and
using the user's selection to refine the composite validation metric by changing one or more of the assigned weighing of the plurality of different scoring or ranking functions such that the analytical techniques responsible for generating the selected analytical result is scored or ranked more highly.

2. The method of claim 1 , wherein the plurality of analytical techniques is a plurality of different clustering techniques and the plurality of corresponding analytical results is a plurality of different clusterings of the same received dataset.

3. The method of claim 1 , wherein the plurality of analytical techniques includes frequent pattern mining techniques, anomaly detection techniques, or factor analysis techniques.

4. The method of claim 1 , wherein the composite validation metric includes elements from each of the plurality of different scoring or ranking functions along with a set of parameters that defines a relative weighing of each element within the composite validation metric.

5. The method of claim 1 , wherein the presenting of the results includes listing the results according to rank order as determined by the composite validation metric.

6. The method of claim 1 , wherein the presenting of the results includes listing the results alongside a score determined by the composite validation metric.

7. The method of claim 1 , wherein using the user's selection to modify the composite validation metric includes employing one or more learning algorithms.

8. The method of claim 1 , additionally comprising:
receiving a second data set;
applying each of the plurality of analytical techniques to the received second data set to generate a second plurality of corresponding analytical results;
applying the refined composite validation metric to each of the second plurality of analytical results; and
displaying the second plurality of analytical results to the user arranged in accordance with the results of the applying the refined composite validation metric to each of the second plurality of analytical results.

9. The method of claim 1 , wherein displaying the plurality of analytical results to the user includes displaying a subset of highest scoring or ranking results.

10. A method for performing assisted knowledge discovery, comprising:
receiving a dataset;
applying each of a plurality of different clustering techniques to the received data set to generate a plurality of corresponding clustering results;
receiving a composite validation metric that is a single scoring or ranking function that is created by combining a plurality of different scoring or ranking functions, each of which is assigned a weighing that determines its relative influence on the composite validation metric;
modifying the received composite validation metric by removing therefrom each of the plurality of different scoring or ranking functions that is assigned a weighing that is less than a predetermined value;
applying the modified composite valuation metric to each of the plurality of analytical results to place the results in an order of importance;
presenting the plurality of clustering results to a user arranged in the order determined by applying the modified composite validation metric;
receiving a selection from the user from among the plurality of clustering results; and
using the user's selection to refine the composition of the composite validation metric by changing one or more of the assigned weighing of the plurality of different scoring or ranking functions.

11. The method of claim 10 , wherein using the user's selection to refine the composition of the composite validation metric includes employing one or more learning algorithms to adapt the composite validation metric such that the clustering techniques responsible for generating the selected analytical result is scored or ranked more highly.

Assignees

• IBM

Inventors

• Biem Alain
• Halim Nagui
• Parthasarathy Srinivasan
• Sow Daby M
• Turaga Deepak S
• Vu Long H

Classifications

• G06N99/005 Primary

  Physics · mapped topic

• G06N20/00 Primary

  Machine learning · CPC title