# US 2022/0092672 A1 · Contextual set selection

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

- Number: US 2022/0092672 A1
- Document ID: US20220092672A1
- Status: Published application
- Kind: A1
- Application number: 17/539,424
- Filed: 2021-12-01
- Published: 2022-03-24
- Earliest claimed: 2019-07-03
- Assignees: Stitch Fix, Inc.
- Applicants: Stitch Fix, Inc.
- Inventors: Kevin J. Zielnicki
- Web page: https://patentgrep.com/patent/US20220092672A1

## CPC

- G06Q 30/0631: {Recommending goods or services}
- G06N 20/00: Machine learning
- G06Q 10/083: Shipping
- 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
- G06N: COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS

## Abstract

An initial list of candidate items automatically evaluated and chosen for an end-user is provided. A selection of one or more items in the initial list of candidate items is received from an expert user different than the end-user to include in an item group set for the end-user. Eligible items are evaluated to identify an additional item to include in the item group set based at least in part on an expert judgement prediction machine learning model trained to predict based at least in part on the one or more items already in the item group set a likelihood of a certain item being evaluated would be selected for inclusion in the item group set. Based on the evaluation, the additional item is included in the item group set. Member items in the item group set are indicated.

## Claims

### Claim 1

A method, comprising:
providing an initial list of candidate items automatically evaluated and chosen for an end-user;
receiving a selection of one or more items in the initial list of candidate items from an expert user different than the end-user to include in an item group set for the end-user;
evaluating eligible items to identify an additional item to include in the item group set based at least in part on an output prediction machine learning model trained to estimate a probability the end-user will like a given item selected by the expert user for the end user and an expert judgment prediction machine learning model trained to predict based at least in part on the one or more items already in the item group set a likelihood of a certain item being evaluated would be selected for inclusion in the item group set, wherein the output prediction machine learning model is trained based in part on previous purchase histories of a plurality of different end-users;
based on the evaluation, including the additional item in the item group set;
indicating member items in the item group set;
receiving end-user feedback regarding the member items in the item group set; and
retraining the output prediction machine learning model based on the end-user feedback.

### Claim 2 (depends on 1)

The method of claim 1, wherein the output prediction machine learning model and the expert judgment prediction machine learning model generate a corresponding overall score for each item included in the item group set.

### Claim 3 (depends on 2)

The method of claim 2, wherein evaluating the initial list of candidate items includes using the expert judgment prediction machine learning model.

### Claim 4 (depends on 1)

The method of claim 1, wherein the initial list of candidate items are automatically ranked and provided in an order based on the rankings.

### Claim 5 (depends on 1)

The method of claim 1, wherein the item group set is to include a preconfigured number of items to be shipped together to the end-user.

### Claim 6 (depends on 1)

The method of claim 1, wherein evaluating the eligible items includes determining a corresponding machine learning prediction score for each of the eligible items among an inventory of items and identifying an item with a best prediction score as the additional item to include in the item group set.

### Claim 7 (depends on 7)

The method of claim 7, wherein determining the machine learning prediction score includes utilizing an outcome prediction machine learning model trained to predict a likelihood that the certain item being evaluated would be purchased by the end-user.

### Claim 8 (depends on 8)

The method of claim 8, wherein determining the machine learning prediction score includes determining a weighted sum of an output of the expert judgment prediction machine learning model and an output of the outcome prediction machine learning model.

### Claim 9 (depends on 1)

The method of claim 1, wherein a plurality of items are to be automatically identified for inclusion in the item group set and each additional item in the plurality of items to be automatically identified is serially determined by using different iterations of the expert judgment prediction machine learning model that accounts for any previously determined additional item in the item group set to determine a next additional item in the item group set.

### Claim 10 (depends on 1)

The method of claim 1, further comprising receiving expert user feedback that includes receiving an approval of the member items in the item group set from the expert user and causing the item group set to be provided to the end-user.

### Claim 11 (depends on 1)

The method of claim 1, wherein the expert user feedback includes receiving an indication from the expert user to replace one of the member items in the item group set.

### Claim 12 (depends on 12)

The method of claim 12, further comprising evaluating at least a portion of the eligible items to identify a ranked list of replacement item candidates at least in part by using the expert judgment prediction machine learning model.

### Claim 13 (depends on 13)

The method of claim 13, wherein the expert user feedback includes receiving a selection of one of the replacement item candidates from the expert user and further comprising replacing the indicated member item in the item group set with the selected replacement item candidate.

### Claim 14 (depends on 1)

The method of claim 1, wherein the item group set includes clothing items.

### Claim 15

A system, comprising:
a processor configured to:
provide an initial list of candidate items automatically evaluated and chosen for an end-user;
receive a selection of one or more items in the initial list of candidate items from an expert user different than the end-user to include in an item group set for the end-user;
evaluate eligible items to identify an additional item to include in the item group set based at least in part on an output prediction machine learning model trained to estimate a probability the end-user will like a given item selected by the expert user for the end user and an expert judgment prediction machine learning model trained to predict based at least in part on the one or more items already in the item group set a likelihood of a certain item being evaluated would be selected for inclusion in the item group set, wherein the output prediction machine learning model is trained based in part on previous purchase histories of a plurality of different end-users;
based on the evaluation, include the additional item in the item group set;
indicate member items in the item group set;
receive end-user feedback regarding the member items in the item group set; and
retrain the output prediction machine learning model based on the end-user feedback; and
a memory coupled to the processor and configured to provide the processor with instructions.

### Claim 16 (depends on 16)

The system of claim 16, wherein the output prediction machine learning model and the expert judgment prediction machine learning model generate a corresponding overall score for each item included in the item group set.

### Claim 17 (depends on 16)

The system of claim 16, wherein evaluating the eligible items includes utilizing the outcome prediction machine learning model trained to predict the likelihood that the certain item being evaluated would be purchased by the end-user.

### Claim 18 (depends on 16)

The system of claim 16, wherein a plurality of items are to be automatically identified for inclusion in the item group set and each additional item in the plurality of items to be automatically identified is serially determined by using different iterations of the expert judgment prediction machine learning model that accounts for any previously determined additional item in the item group set to determine a next additional item in the item group set.

### Claim 19 (depends on 16)

The system of claim 16, wherein the feedback includes an indication from the expert user to replace one of the member items in the item group set and the processor is further configured to evaluate at least a portion of the eligible items to identify a ranked list of replacement item candidates at least in part by using the expert judgment prediction machine learning model.

### Claim 20

A computer program product, the computer program product being embodied in a non-transitory computer readable storage medium and comprising computer instructions for:
providing an initial list of candidate items automatically evaluated and chosen for an end-user;
receiving a selection of one or more items in the initial list of candidate items from an expert user different than the end-user to include in an item group set for the end-user;
evaluating eligible items to identify an additional item to include in the item group set based at least in part on an output prediction machine learning model trained to estimate a probability the end-user will like a given item selected by the expert user for the end user and an expert judgment prediction machine learning model trained to predict based at least in part on the one or more items already in the item group set a likelihood of a certain item being evaluated would be selected for inclusion in the item group set, wherein the output prediction machine learning model is trained based in part on previous purchase histories of a plurality of different end-users;
based on the evaluation, including the additional item in the item group set;
indicating member items in the item group set;
receiving end-user feedback regarding the member items in the item group set; and
retraining the output prediction machine learning model based on the end-user feedback.
