# US 12,008,587 B2 · Methods and apparatus to generate audience metrics using matrix analysis

> Agents: this data is available as JSON from the patentgrep API and MCP server — see https://patentgrep.com/llms.txt

- Number: US 12,008,587 B2
- Document ID: US12008587B2
- Status: Granted
- Kind: B2
- Application number: 17/406,878
- Filed: 2021-08-19
- Published: 2022-02-24
- Granted: 2024-06-11
- Earliest claimed: 2020-08-21
- Assignees: The Nielsen Company (US), LLC
- Applicants: The Nielsen Company (US), LLC
- Inventors: DongBo Cui; Edward Murphy; Michael Richard Sheppard; David Forteguerre; Jessica Lynn White; PengFei Yi; Jonathan Sullivan
- Web page: https://patentgrep.com/patent/US12008587B2

## CPC

- G06Q 30/0201: Market modelling; Market analysis; Collecting market data
- G06F 17/16: Matrix or vector computation {, e.g. matrix-matrix or matrix-vector multiplication, matrix factorization (matrix transposition G06F7/78)}
- G06F 18/10: Pre-processing; Data cleansing
- G06V 10/764: using classification, e.g. of video objects
- G06Q: INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES; SYSTEMS OR METHODS SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES, NOT OTHERWISE PROVIDED FOR
- G06F: ELECTRIC DIGITAL DATA PROCESSING (computer systems based on specific computational models G06N)
- G06V: IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING

## Abstract

An example apparatus includes audience metrics collecting circuitry to access first audience metrics from a server, and access second audience metrics from the server, matrix building circuitry to build a matrix of the first audience metrics and the second audience metrics, missing values of the matrix corresponding to the second audience metrics, data transforming circuitry to transform the first audience metrics and the second audience metrics in the matrix, missing value calculating circuitry to determine imputed transformed values of the missing values using a recommender system, and the data transforming circuitry to recover imputed values of the missing values based on the imputed transformed values.

## Claims

### Claim 1

An audience measurement system comprising:
a panel monitor server configured to:
receive network communications identifying media items presented at client devices of panelists of an audience measurement entity, wherein the media items are tagged with beacon instructions that are downloaded to the client devices when the client devices access the media items, the beacon instructions to cause the client devices to transmit the network communications to the panel monitor server,
based on the network communications received from the client devices in accordance with the beacon instructions, log demographic impressions of the media items in association with known demographic data for the panelists, and
generate panelist audience metrics based on the demographic impressions;
a computing system comprising at least on processor and a memory, the computing system configured to:
access the panelist audience metrics from the panel monitor server,
access, from a census monitor server, census audience metrics generated based on census impressions of the media items logged by the census monitor server, wherein demographic data for users that accessed the media items was not logged by the census monitor server in association with the census impressions,
build a sparse matrix of the panelist audience metrics and the census audience metrics, wherein the sparse matrix comprises missing audience metrics values for a portion of demographic groups and for a portion of the media items, and wherein the sparse matrix comprises computer-generated audience metrics data bias represented by the missing audience metrics values,
normalize the panelist audience metrics and the census audience metrics in the sparse matrix to generate a sparse transformed matrix,
apply a recommender model to the sparse transformed matrix to predict audience metrics values corresponding to the missing audience metrics,
generate a supplemented transformed matrix comprising the predicted audience metrics values and audience metrics values from the sparse transformed matrix,
perform reverse transformations of the supplemented transformed matrix to generate a completed audience metrics matrix, and
transmit audience data to the panel monitor server and the census monitor server to cause the panel monitor server and the census monitor server to update respective audience metrics databases.

### Claim 2 (depends on 1)

The audience measurement system of claim 1, wherein the panelist audience metrics include a first audience size, a first impression count, and first duration data, and the census audience metrics include a second audience size, a second impression count, and second duration data.

### Claim 3 (depends on 1)

The audience measurement system of claim 1, wherein normalizing the panelist audience metrics and the census audience metrics in the sparse matrix to generate a sparse transformed matrix comprises:
normalizing the panelist audience metrics and the census audience metrics to a universe estimate;
logically transforming the normalized metrics to generate logically transformed metrics; and
performing a logarithmic transformation of the logically transformed metrics to generate the sparse transformed matrix.

### Claim 4 (depends on 1)

The audience measurement system of claim 1, wherein the recommender model comprises a deep factorization machine.

### Claim 5 (depends on 1)

The audience measurement system of claim 1, wherein the recommender model comprises a multilayer neural network.

### Claim 6 (depends on 1)

The audience measurement system of claim 1, wherein the media items are streaming media items.

### Claim 7 (depends on 1)

The audience measurement system of claim 1, further comprising the census monitor server.

### Claim 8

At least one non-transitory computer readable storage medium comprising instructions that, when executed, cause at least one processor to perform operations comprising:
receiving, at a panel monitor server, network communications identifying media items presented at client devices of panelists of an audience measurement entity, wherein the media items are tagged with beacon instructions that are downloaded to the client devices when the client devices access the media items, the beacon instructions to cause the client devices to transmit the network communications to the panel monitor server;
based on the network communications received from the client devices in accordance with the beacon instructions, logging demographic impressions of the media items in association with known demographic data for the panelists;
generating panelist audience metrics based on the demographic impressions;
accessing the panelist audience metrics from the panel monitor server;
accessing, from a census monitor server, census audience metrics generated based on census impressions of the media items logged by the census monitor server, wherein demographic data for users that accessed the media items was not logged by the census monitor server in association with the census impressions;
building a sparse matrix of the panelist audience metrics and the census audience metrics, wherein the spare matrix comprises missing audience metrics values for a portion of demographic groups and or a portion of the media items, and wherein the sparse matrix comprises computer-generated audience metrics data bias represented by the missing audience metrics values;
normalizing the panelist audience metrics and the census audience metrics in the sparse matrix to generate a sparse transformed matrix;
applying a recommender model to the sparse transformed matrix to predict audience metrics values corresponding to the missing audience metrics;
generating a supplemented transformed matrix comprising the predicted audience metrics values and audience metrics values from the sparse transformed matrix;
performing reverse transformations of the supplemented transformed matrix to generate a completed audience metrics matrix; and
transmitting audience metrics data to the panel monitor server and the census monitor server to cause the panel monitor server and the census monitor server to update respective audience metrics databases.

### Claim 9 (depends on 8)

The at least one non-transitory computer readable storage medium of claim 8, wherein the panelist audience metrics include a first audience size, a first impression count, and first duration data, and the census audience metrics include a second audience size, a second impression count, and second duration data.

### Claim 10 (depends on 8)

The at least one non-transitory computer readable storage medium of claim 8, wherein normalizing the panelist audience metrics and the census audience metrics in the sparse matrix to generate a sparse transformed matrix comprises:
normalizing the panelist audience metrics and the census audience metrics to a universe estimate;
logically transforming the normalized metrics to generate logically transformed metrics; and
performing a logarithmic transformation of the logically transformed metrics to generate the sparse transformed matrix.

### Claim 11 (depends on 8)

The at least one non-transitory computer readable storage medium of claim 8, wherein the recommender model comprises a deep factorization machine.

### Claim 12 (depends on 8)

The at least one non-transitory computer readable storage medium of claim 8, wherein the recommender model comprises a multilayer neural network.

### Claim 13 (depends on 8)

The at least one non-transitory computer readable storage medium of claim 8, wherein the media items are streaming media items.

### Claim 14 (depends on 8)

The at least one non-transitory computer readable storage medium of claim 8, wherein the client devices comprise smartphones.

### Claim 15

A method comprising:
receiving, by a panel monitor server, network communications identifying media items presented at client devices of panelists of an audience measurement entity, wherein the media items are tagged with beacon instructions that are downloaded to the client devices when the client devices access the media items, the beacon instructions to cause the client devices to transmit the network communications to the panel monitor server;
based on the network communications received from the client devices in accordance with the beacon instructions, logging, by the panel monitor server, demographic impressions of the media items in association with known demographic data for the panelists;
generating, by the panel monitor server, panelist audience metrics based on the demographic impressions;
accessing, by a computing system different from the panel monitor server, the panelist audience metrics from the panel monitor server;
accessing, by the computing system, from a census monitor server, census audience metrics generated based on census impressions of the media items logged by the census monitor server, wherein demographic data for users that accessed the media items was not logged by the census monitor server in association with the census impressions;
building, by the computing system, a sparse matrix of the panelist audience metrics and the census audience metrics, wherein the spare matrix comprises missing audience metrics values for a portion of demographic groups and or a portion of the media items, and wherein the sparse matrix comprises computer-generated audience metrics data bias represented by the missing audience metrics values;
normalizing, by the computing system, the panelist audience metrics and the census audience metrics in the sparse matrix to generate a sparse transformed matrix;
applying, by the computing system, a recommender model to the sparse transformed matrix to predict audience metrics values corresponding to the missing audience metrics;
generating, by the computing system, a supplemented transformed matrix comprising the predicted audience metrics values and audience metrics values from the sparse transformed matrix;
performing by the computing system, reverse transformations of the supplemented transformed matrix to generate a completed audience metrics matrix; and
transmitting by the computing system, audience metrics data to the panel monitor server and the census monitor server to cause the panel monitor server and the census monitor server to update respective audience metrics databases.

### Claim 16 (depends on 15)

The method of claim 15, wherein normalizing the panelist audience metrics and the census audience metrics in the sparse matrix to generate a sparse transformed matrix comprises:
normalizing the panelist audience metrics and the census audience metrics to a universe estimate;
logically transforming the normalized metrics to generate logically transformed metrics; and
performing a logarithmic transformation of the logically transformed metrics to generate the sparse transformed matrix.

### Claim 17 (depends on 15)

The method of claim 15, wherein the recommender model comprises a deep factorization machine.

### Claim 18 (depends on 15)

The method of claim 15, wherein the recommender model comprises a multilayer neural network.

### Claim 19 (depends on 15)

The method of claim 15, wherein the media items are streaming media items.

### Claim 20 (depends on 15)

The method of claim 15, wherein the panelist audience metrics include a first audience size, a first impression count, and first duration data, and the census audience metrics include a second audience size, a second impression count, and second duration data.