Next location prediction

Patent metadata

Field	Value
Publication number	US-11200403-B2
Application number	US-201615141484-A
Country	US
Kind code	B2
Filing date	Apr 28, 2016
Priority date	Apr 28, 2016
Publication date	Dec 14, 2021
Grant date	Dec 14, 2021

Abstract

A method and system are provided for next location prediction. The method includes inferring, by a hardware processor, a store layout, based on user location data and user transaction data for a plurality of users. The method further includes clustering, by the hardware processor, the plurality of users based on the user transaction data to form a set of clusters. The method also includes ensembling, by the hardware processor, users within each of the clusters and building a location prediction model for each of the clusters. The method additionally includes predicting, by the hardware processor, a next location of a particular user from the plurality of users based on a weighted vote taken over the location prediction model for the cluster corresponding to the particular user. The cluster corresponding to the particular user includes at least one other user from the plurality of users in addition to the particular user.

First claim

What is claimed is:

1. A method for next location prediction, comprising:
recognizing, by a facial recognition system, a plurality of users entering a retail store to transform facial features of the plurality of users into user identifiers;
clustering, by a hardware processor, the plurality of users recognized by the facial recognition system based on user transaction data and the user identifiers to form a set of clusters;
ensembling, by the hardware processor, the plurality of users within clusters in a set of clusters formed based on user transaction data, and building a next location prediction model for each of the clusters to form a next location prediction model ensemble by converting a probability distribution over products of interest into a probability distribution over locations the plurality of users in respective ones of the clusters are likely interested in visiting using a store layout, semantic labels, and a respective Markov chain built over the plurality of users in each of the respective ones of the clusters;
predicting, by the hardware processor, a next location of a particular user from the plurality of users in a same cluster based on a weighted vote performed over a combined distribution of next locations predicted by the next location prediction model ensemble for the cluster corresponding to the particular user and next locations predicted by a customer intent model; and
directing the particular user to an item determined to be of interest to the particular user, based on the predicted next location of the user,
wherein the cluster corresponding to the particular user comprises at least one other user from the plurality of users in addition to the particular user.

2. The method of claim 1 , wherein the item has an in-store location within a threshold distance to the predicted next location of the particular user.

3. The method of claim 1 , further comprising:
semantically enriching, by the hardware processor, user location data with product information for products located proximate to locations indicated by the user location data; and
displaying, on a display device, the product information to the user when the user is at the locations proximate to the products.

4. The method of claim 3 , wherein the next location prediction model for each of the clusters is built based on semantic labels determined from said enriching step.

5. The method of claim 1 , further comprising:
determining at least one store product of interest to the particular user and an in-store location of the at least one store product; and
directing the particular user to the at least one store product at the in-store location responsive to the particular user entering a store or becoming within a threshold distance of the at least one store product.

6. The method of claim 1 , further comprising preserving an individual privacy of the plurality of users by combining user movement patterns across at least multiple ones of the plurality of users.

7. The method of claim 1 , wherein user location data comprises a timestamp, a location identifier, and a user identifier.

8. The method of claim 1 , wherein the user transaction data comprises a timestamp, a purchased product type, and a user identifier.

9. The method of claim 1 , wherein the next location prediction model is configured to delineate weekend next location predictions versus weekday next location predictions.

10. The method of claim 1 , wherein the next location prediction for the particular customer is based on past model states of the next location prediction model for the cluster corresponding to the particular customer.

11. The method of claim 1 , transmitting, to a mobile device carried by the particular user, a coupon for an item proximate to the predicted next location.

12. The method of claim 1 , further comprising protecting a privacy of a specific one of the plurality of users by generating the predicted next location of the user to be non-user-specific and cluster-specific.

13. The method of claim 1 , wherein the customer intent model produces a probability distribution over a set of items that the particular user is interested in.

14. A computer program product for next location prediction, the computer program product comprising a non-transitory computer readable storage medium having program instructions embodied therewith, the program instructions executable by a computer to cause the computer to perform a method comprising:
recognizing, by a facial recognition system, a plurality of users entering a retail store to transform facial features of the plurality of users into user identifiers;
clustering, by a hardware processor, the plurality of users recognized by the facial recognition system based on user transaction data and the user identifiers to form a set of clusters;
ensembling, by the hardware processor, the plurality of users within clusters in a set of clusters formed based on user transaction data, and building a next location prediction model for each of the clusters to form a next location prediction model ensemble by converting a probability distribution over products of interest into a probability distribution over locations the plurality of users in respective ones of the clusters are likely interested in visiting using a store layout, semantic labels, and a respective Markov chain built over the plurality of users in each of the respective ones of the clusters;
predicting, by the hardware processor, a next location of a particular user from the plurality of users in a same cluster based on a weighted vote performed over a combined distribution of next locations predicted by the next location prediction model ensemble for the cluster corresponding to the particular user and next locations predicted by a customer intent model; and
directing the particular user to an item determined to be of interest to the particular user, based on the predicted next location of the user by the next location prediction model and a predicted next location by a customer intent model,
wherein the cluster corresponding to the particular user comprises at least one other user from the plurality of users in addition to the particular user.

15. A system for next location prediction, comprising:
a facial recognition system for recognizing a plurality of customers responsive to the plurality of users entering a retail store to transform facial features of the plurality of users into user identifiers; and
a hardware processor, configured to:
cluster the plurality of users recognized by the facial recognition system based on user transaction data and the user identifiers to form a set of clusters;
ensemble the plurality of users within clusters in a set of clusters formed based on user transaction data and building a next location prediction model for each of the clusters to form a next location prediction model ensemble by converting a probability distribution over products of interest into a probability distribution over locations the plurality of users in respective ones of the clusters are likely interested in visiting using a store layout, semantic labels, and a respective Markov chain built over the plurality of users in each of the respective ones of the clusters;
predicting, by the hardware processor, a next location of a particular user from the plurality of users in a same cluster based on a weighted vote performed over a combined distribution of next locations predicted by the next location prediction model ensemble for the cluster corresponding to the particular user and next locations predicted by a customer intent model; and
direct the pa

Assignees

• IBM

Inventors

• Ganti Raghu K
• Herman Norbert
• Ray Shubhadip
• Srivatsa Mudhakar

Classifications

• H04W4/029 Primary

  Location-based management or tracking services · CPC title

• G06V10/762

  using clustering, e.g. of similar faces in social networks · CPC title

• G06V10/764

  using classification, e.g. of video objects · CPC title

• G06V40/16

  Human faces, e.g. facial parts, sketches or expressions · CPC title

• G06F18/23

  Clustering techniques · CPC title