MACHINE-BASED PREDICTION OF VISITATION CAUSED BY VIEWING

• US 20200134420A1
• Filed: 10/25/2018
• Published: 04/30/2020
• Est. Priority Date: 10/25/2018
• Status: Active Grant

• Current Assignee: Billups, Inc.
• Original Assignee: Boohma Technologies, LLC
• Inventors: Spooner, Shawn
• Granted Patent: US 11,694,063 B2
• Time in Patent Office: Days
• CPC Class Codes: G06F 16/29 Geographical information databases
• CPC Class Codes: G06F 16/9024 Graphs; Linked lists G06F16/9027 takes precedence
• CPC Class Codes: G06F 16/907 Retrieval characterised by using metadata, e.g. metadata not derived from the content or metadata generated manually
• CPC Class Codes: G06F 17/18 for evaluating statistical data , e.g. average values, frequency distributions, probability functions, regression analysis forecasting specially adapted for a specific administrative, business or logistic context G06Q10/04
• CPC Class Codes: G06N 3/045 Combinations of networks
• CPC Class Codes: G06N 3/08 Learning methods
• CPC Class Codes: G06N 5/01 Dynamic search techniques; Heuristics; Dynamic trees; Branch-and-bound
• CPC Class Codes: G06N 7/01 Probabilistic graphical models, e.g. probabilistic networks
• CPC Class Codes: H04W 4/02 Services making use of location information
• CPC Class Codes: H04W 4/023 using mutual or relative location information between multiple location based services [LBS] targets or of distance thresholds
• CPC Class Codes: H04W 4/024 Guidance services
• CPC Class Codes: H04W 4/029 Location-based management or tracking services

Abstract

For machine-based prediction of visitation, a graph of paired mobile devices is formed where the edges for each pair are based on similarities in visitation and/or metadata for the devices. A machine-learned network embeds the visitation and metadata information, which is used to indicate the similarity. Since the trace data used to show access may be sparse, another machine-learned network completes the route based on routes used by similar devices. Another machine-learned network recommends effectiveness of content based on routes, the graph, metadata, and/or other information. The recommendation is based on training using counterfactual and/or other causal modeling.

Claims

1. A method for machine-based prediction of visitation, the method comprising:

acquiring first trace data showing routes of devices over time in a region, the first trace data including unobserved portions for at least some of the devices;

forming, by a second machine-learned network, a first graph linking pairs of the devices with edges assigned a similarity between the pairs of the devices;

generating, by a first machine-learned network trained using deep learning, second trace data for the unobserved portions for the at least some of the devices;

modeling, by a third machine-learned network, a causal effect of first content to visits of a location from the first graph; and

displaying a prediction of effectiveness of second content based on the causal effect.

2. The method of claim 1 wherein generating comprises generating with the first machine-learned network comprising a deep Markov model.

3. The method of claim 1 further comprising dividing the devices into behavior cohorts, and wherein generating comprises generating, for each device, the second trace data based on patterns from the behavior cohort to which the device is a member.

4. The method of claim 3 wherein dividing comprises dividing based on the first graph.

5. The method of claim 1 wherein forming comprises forming with the second machine-learned network comprising a deep learned neural network.

6. The method of claim 1 wherein forming comprises forming with graph embeddings of similarity in frequency and location for the first and second trace data and in metadata related to behavior linked to the devices.

7. The method of claim 6 wherein the graph embeddings form a lower dimensional vector from the frequency, location, and metadata.

8. The method of claim 1 wherein modeling comprising modeling with the third machine-learned network trained by determining a first sub-set of other devices exposed to first content of a viewable surface during a period and a second sub-set of the other devices not exposed to the first content of the viewable surface.

9. The method of claim 8 wherein determining comprises determining with an aggregation of based on the first and second trace data and based on another source.

10. The method of claim 1 wherein modeling comprises modeling the casual effect with the third machine-learned network trained with a counter factual probability.

11. The method of claim 10 wherein modeling comprises modeling with the third machine-learned network trained with minimization of distances between pairs of the devices based on the similarities and identifying a likelihood given the first sub-set not being exposed to the first content.

12. The method of claim 1 wherein modeling comprises modeling with the third machine-learned network trained with a probability of visitation of the location based on the first and second trace data and an embedding from the forming of the first graph for before the period.

13. The method of claim 1 wherein modeling comprises modeling a probability of viewing.

14. The method of claim 1 further comprising identifying covariates of the first graph and predicting the effectiveness, wherein displaying comprises displaying the covariates and the effectiveness.

15. The method of claim 1 wherein displaying comprises displaying causal relationships in relation to the second content, timing of the second content, and location of the second content.

16. A method for machine training of prediction of visitation, the method comprising:

assigning devices for a pre-treatment period as having visited or not visited a first location;

assigning the devices for a treatment period as having visited or not visited the first location;

identifying pairs of devices based on times and frequencies of visitation of second locations;

connecting the pairs of devices by a similarity from a machine-learned neural network, the similarity based on the times and frequencies;

minimizing a distance based on the similarity of the pairs between the pre-treatment period and the treatment period with counterfactual reasoning;

determining an effect of treatment based on the distance; and

machine training a neural network based on the effect of the treatment.

17. The method of claim 16 wherein identifying and connecting comprises forming a graph, and wherein minimizing comprises minimizing for greatest similarity in the graph.

18. The method of claim 17 further comprising reducing the similarities for most similar pairs.

19. The method of claim 16 wherein minimizing comprises minimizing as a function of behavioral metadata for the pairs and visitation and travel pattern histories of the pairs.

20. The method of claim 16 further comprising predicting effect of visitation patterns of the devices based on the effect of the treatment.