# US 2019/0244253 A1 · Target identification using big data and machine learning

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

- Number: US 2019/0244253 A1
- Document ID: US20190244253A1
- Status: Published application
- Kind: A1
- Application number: 16/046,507
- Filed: 2018-07-26
- Published: 2019-08-08
- Earliest claimed: 2018-02-06
- Assignees: Accenture Global Solutions Limited
- Applicants: Accenture Global Solutions Limited
- Inventors: Abhinav Vij; Gaurav Goyal; Prithika Priyanshi; Anupam Shringi; Chinmay Jha; Saurabh Aggarwal; Abhishek Sunnak; Vijay Saini; Aashish Kumar; Shivam Singh Baghel; Rashmi Aggarwal
- Web page: https://patentgrep.com/patent/US20190244253A1

## CPC

- G06Q 30/0271: {Personalized advertisement}
- G06F 18/211: Selection of the most significant subset of features
- G06F 18/2135: based on approximation criteria, e.g. principal component analysis
- G06F 18/214: Generating training patterns; Bootstrap methods, e.g. bagging or boosting
- G06F 18/23213: with fixed number of clusters, e.g. K-means clustering
- G06F 18/2411: based on the proximity to a decision surface, e.g. support vector machines
- G06F 18/24143: {Distances to neighbourhood prototypes, e.g. restricted Coulomb energy networks [RCEN]}
- G06F 40/00: Handling natural language data (speech analysis or synthesis, speech recognition G10L)
- G06F 40/30: Semantic analysis
- G06N 20/00: Machine learning
- G06N 20/20: Ensemble learning
- G06Q 30/0276: {Advertisement creation}
- G06N 5/01: Dynamic search techniques; Heuristics; Dynamic trees; Branch-and-bound
- G06N 7/01: Probabilistic graphical models, e.g. probabilistic networks
- 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)
- G06N: COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS

## Abstract

A device may obtain, from a collection of data sources, personal information and activity information for a group of individuals. The device may generate profile information by associating the personal information and the activity information for each individual in the group of individuals. The device may determine a set of features capable of being used to train a set of machine learning models by using one or more feature identification techniques to analyze the profile information. The set of features may include generic features and contextualized features associated with the group of individuals. The device may train the set of machine learning models using one or more subsets of features of the set of features. The device may use the set of machine learning models to process a request from a client device for a list of prospective targets for a campaign.

## Claims

### Claim 1

A device, comprising:
one or more memories; and
one or more processors, operatively connected to the one or more memories, to:
obtain, from a collection of data sources, personal information and activity information for a group of individuals;
generate profile information by associating the personal information and the activity information for each individual in the group of individuals;
determine a set of features capable of being used to train a set of machine learning models by using one or more feature identification techniques to analyze the profile information,
the set of features including generic features and contextualized features associated with the group of individuals;
select, using one or more feature selection techniques, one or more subsets of features, of the set of features, to be used for training the set of machine learning models,
where each subset of features, of the one or more subsets of features, identifies measurable properties associated with one or more segments of individuals within the group of individuals;
train the set of machine learning models using the one or more subsets of features,
the set of machine learning models to be trained to output values indicating a degree to which each individual in the group of individuals is a prospective target for a campaign;
receive, from a client device, a request for a list of prospective targets to target for the campaign;
use the set of machine learning models to determine the list of prospective targets for the campaign; and
provide information associated with the list of prospective targets to the client device to allow the list of prospective targets to be used in the campaign.

### Claim 2 (depends on 1)

The device of claim 1, where the one or more processors, when generating the profile information, are to:
standardize the personal information and the activity information into a uniform format,
associate personal information values, associated with the personal information, with activity information values, associated with the activity information, using one or more natural language processing techniques,
the one or more natural language processing techniques including a fuzzy matching technique to associate a first one or more personal information values that are not direct matches with a first one or more activity information values, and
generate the profile information for each individual in the group of individuals based on associating the personal information values with the activity information values.

### Claim 3 (depends on 1)

The device of claim 1, where the one or more processors, when determining the set of features, are to:
determine, by using a feature identification technique, of the one or more feature identification techniques, one or more features, of the set of features, that are based on combinations of profile information values,
wherein the combinations of profile information values are indicators of a particular individual's or segment of individuals' propensity to make a purchasing decision associated with the campaign.

### Claim 4 (depends on 1)

The device of claim 1, where the one or more processors, when selecting the subset of features, are to:
determine that profile information values for a first segment of individuals includes a threshold amount of non-null values, and
select a subset of contextualized features to apply to the profile information values of the first segment of individuals.

### Claim 5 (depends on 1)

The device of claim 1, where the one or more processors, when selecting the one or more subsets of features, are to:
determine that profile information values for a first segment of individuals include a threshold amount of null values,
select, for the profile information values for the first segment of individuals, a first subset of generalized features using at least one of:
a variable selection using random forest (VSURF) analysis,
a bootstrapping analysis, or
a regularization analysis; and
where the one or more processors, when training the set of machine learning models, are to:
train a machine learning model, of the set of machine learning models, using the first subset of generalized features.

### Claim 6 (depends on 1)

The device of claim 1, where the one or more processors, when selecting the subset of features, are to:
select a subset of contextualized features of the contextualized features,
the subset of contextualized features being associated with individuals that performed an affirmative action toward a purchase of a product or a service; and
where the one or more processors, when training the set of machine learning models, are to:
train a machine learning model, of the set of machine learning models, using the subset of contextualized features.

### Claim 7 (depends on 1)

The device of claim 1, where the one or more processors are further to:
generate, after using the set of machine learning models to determine the list of prospective targets, a recommendation indicating how to deploy the campaign,
the recommendation including at least one of:
an instruction indicating a medium to use when influencing a particular individual that is included in the list of prospective targets,
an instruction indicating a time of day during which the particular individual is most likely to be online to view a particular advertisement, or
an instruction indicating to make the campaign available to a particular type of device that is a preferred device of the particular individual; and
where the one or more processors, when providing the information associated with the list of prospective targets to the client device, are to:
provide the recommendation to the client device.

### Claim 8

A non-transitory computer-readable medium storing instructions, the instructions comprising:
one or more instructions that, when executed by one or more processors, cause the one or more processors to:
receive a set of machine learning models that have been trained using one or more subsets of features, of a set of features that include generic features and contextualized features associated with a group of individuals,
wherein each subset of features, of the one or more subsets of features, identify measurable properties associated with one or more segments of individuals within the group of individuals, and
wherein the set of machine learning models are trained to output values indicating a degree to which each individual in the group of individuals is a prospective target for a campaign;
receive, from a client device, a request for a list of prospective targets for the campaign;
use the set of machine learning models to determine the list of prospective targets to use for the campaign; and
provide information associated with the list of prospective targets to the client device to allow the list of prospective targets to be used in the campaign.

### Claim 9 (depends on 8)

The non-transitory computer-readable medium of claim 8, where the one or more instructions, that cause the one or more processors to receive the set of machine learning models, cause the one or more processors to:
receive the set of machine learning models,
wherein a machine learning model, of the set of machine learning models, has been trained on a subset of the contextualized features, and
wherein the subset of the contextualized features is associated with individuals that performed an affirmative action toward a purchase of a product or a service.

### Claim 10 (depends on 8)

The non-transitory computer-readable medium of claim 8, where the one or more instructions, that cause the one or more processors to receive the set of machine learning models, cause the one or more processors to:
receive the set of machine learning models,
wherein a machine learning model, of the set of machine learning models, has been trained on a subset of the contextualized features, and
wherein the machine learning model has been trained on the subset of the contextualized features based on profile information values associated with one or more data sources satisfying a threshold amount of non-null values.

### Claim 11 (depends on 8)

The non-transitory computer-readable medium of claim 8, where the one or more instructions, that cause the one or more processors to receive the set of machine learning models, cause the one or more processors to:
receive the set of machine learning models,
wherein a machine learning model, of the set of machine learning models, has been trained on a subset of the generic features, and
wherein the machine learning model has been trained on the subset of the generic features based on profile information values associated with one or more data sources failing to satisfy a threshold amount of non-null values.

### Claim 12 (depends on 8)

The non-transitory computer-readable medium of claim 8, where the one or more instructions, that cause the one or more processors to receive the set of machine learning models, cause the one or more processors to:
receive the set of machine learning models,
wherein a machine learning model, of the set of machine learning models, has been trained using features that are based on combinations of profile information values, and
wherein the combinations of profile information values are indicators of a particular individual's or segment of individuals' propensity to make a purchasing decision associated with the campaign.

### Claim 13 (depends on 8)

The non-transitory computer-readable medium of claim 8, where the one or more instructions, when executed by the one or more processors, further cause the one or more processors to:
receive, after providing information associated with the list of prospective targets to the client device, user feedback information indicating a level of effectiveness of the campaign,
modify a machine learning model, of the set of machine learning models, based on the user feedback information,
the machine learning model having been trained using a first one or more subsets of features,
wherein the one or more instructions, that cause the one or more processors to modify the machine learning model, cause the one or more processors to:
remove a feature, of the first one or more subsets of features, from the machine learning model,
add a feature to the first one or more subsets of features of the machine learning model, or
modify a feature of the first one or more subsets of features used in the machine learning model.

### Claim 14

A method, comprising:
obtaining, by a device and from a collection of data sources, personal information and activity information for a group of individuals;
generating, by the device, profile information by associating the personal information and the activity information for each individual in the group of individuals;
determining, by the device, a set of features capable of being used to train a set of machine learning models by using one or more feature identification techniques to analyze the profile information,
the set of features including generic features and contextualized features associated with the group of individuals;
training, by the device, the set of machine learning models using one or more subsets of features of the set of features,
the set of machine learning models to be trained to output values indicating a degree to which each individual in the group of individuals is a prospective target for a campaign; and
using, by the device, the set of machine learning models to process a request from a client device for a list of prospective targets for the campaign.

### Claim 15 (depends on 14)

The method of claim 14, further comprising:
selecting, using one or more feature selection techniques, the one or more subsets of features, of the set of features, to be used for training the set of machine learning models,
where each subset of features identify measurable properties associated with one or more segments of individuals within the group of individuals.

### Claim 16 (depends on 14)

The method of claim 14, further comprising:
determining, after determining the set of features, that profile information values for a first segment of individuals, in the group of individuals, include a threshold amount of null values;
selecting a subset of generic features to apply to the profile information values for the first segment of individuals;
determining that profile information values for a second segment of individuals, in the group of individuals, does not include the threshold amount of null values; and
selecting a subset of contextualized features to apply to the profile information values of the second segment of individuals.

### Claim 17 (depends on 14)

The method of claim 14, further comprising:
selecting a subset of the contextualized features,
the subset of the contextualized features being associated with individuals that performed an affirmative action toward a purchase of a product or a service; and
where training the set of machine learning models comprises:
training a machine learning model, of the set of machine learning models, using the subset of the contextualized features.

### Claim 18 (depends on 14)

The method of claim 14, where generating the profile information comprises:
standardizing the personal information and the activity information into a uniform format,
associating personal information values, associated with the personal information, with activity information values, associated with the activity information, using one or more natural language processing techniques, and
generating the profile information for each individual in the group of individuals based on associating the personal information values with the activity information values.

### Claim 19 (depends on 14)

The method of claim 14, where training the set of machine learning models using the one or more subsets of features comprises:
selecting a first subset of features associated with profile information values obtained from a first data source and a second subset of features associated with profile information values obtained from a second data source,
determining that the first subset of features and the second subset of features satisfy a threshold level of similarity, and
training a first machine learning model, of the set of machine learning models, using the first subset of features and the second subset of features based on determining that the first subset of features and the second subset of features satisfy the threshold level of similarity.

### Claim 20 (depends on 14)

The method of claim 14, where using the set of machine learning models to process the request for the list of prospective targets comprises:
receiving, from the client device, the request for the list of prospective targets for the campaign,
using information included in the request to identify a subset of the group of individuals as candidate prospective targets,
providing profile information for the candidate prospective targets as input to the set of machine learning models to cause the set of machine learning models to output, for the candidate prospective targets, values identifying a predicted level of interest in a good or a service associated with the campaign,
determining whether the values satisfy a threshold,
identifying the list of prospective targets based on determining whether the values satisfy the threshold, and
providing the list of prospective targets and information associated with the list of prospective targets to the client device to allow the list of prospective targets to be used in the campaign.