Estimating shipping costs with machine learning

Patent metadata

Field	Value
Publication number	US-11544661-B2
Application number	US-202016878020-A
Country	US
Kind code	B2
Filing date	May 19, 2020
Priority date	Apr 9, 2020
Publication date	Jan 3, 2023
Grant date	Jan 3, 2023

Abstract

Disclosed herein are system, method, and computer program product embodiments for estimating shipping costs of items purchased in an online market using machine learning techniques. By determining in real-time the dimensional weight of an item with reference to a machine learning model describing past transactions, a buyer and seller in the online market can finalize the transaction in real-time. The machine learning model further implements a bias within the machine learning algorithm towards heavier estimation to avoid undercharging the market participants for shipping costs. One embodiment involves using these cost-estimation techniques in the context of a local shipping feature, which allows buyers and sellers to schedule a same-day delivery by seamlessly involving a local carrier.

First claim

What is claimed is:

1. A computer-implemented method, comprising:
estimating, by one or more processors, a dimensional weight for an item purchased by a first user in an online market, wherein the item is shipped to a destination associated with the first user from an origin associated with a second user using a carrier, wherein the item is associated with a plurality of characteristics, and wherein the dimensional weight is determined by referencing a plurality of learned vectors each representing a previously shipped item;
creating, by the one or more processors, a vector that represents the item based on the plurality of characteristics;
determining, by the one or more processors, one or more nearest vectors among the plurality of learned vectors by calculating distances between the vector that represents the item and the plurality of learned vectors to determine the one or more nearest vectors, wherein each vector in the one or more nearest vectors comprises an actual dimensional weight;
determining, by the one or more processors, the dimensional weight of the item purchased by the first user, by calculating a weighted average of the actual dimensional weight for each vector in the one or more nearest vectors, wherein the actual dimensional weight for a first set of the one or more nearest vectors is adjusted up for heavier weights;
determining, by the one or more processors, an estimated cost based on the dimensional weight of the item and the carrier; and
scheduling, by the one or more processors, a delivery of the item with the carrier for the estimated cost.

2. The method of claim 1 , wherein the creating is performed using a deep learning model comprising an item similarity model.

3. The method of claim 1 , wherein the calculating the weighted average employs a cost function that biases the weighted average towards a heavier estimate.

4. The method of claim 1 , wherein the carrier is a local carrier.

5. The method of claim 1 , the determining the estimated cost further comprising:
sending, by the one or more processors, a query comprising the dimensional weight, the origin, and the destination to an application programming interface offered by the carrier and receiving the estimated cost in response.

6. The method of claim 1 , the scheduling the delivery further comprising:
sending, by the one or more processors, the estimated cost to the second user;
receiving, by the one or more processors, a confirmation of the estimated cost, the origin and a pickup time from the second user;
receiving, by the one or more processors, a confirmation of the destination and a drop-off time from the first user;
accessing, by the one or more processors, an application programming interface offered by the carrier to place a shipping order comprising the item, the dimensional weight, the origin, the pickup time, the destination, and the drop-off time.

7. The method of claim 1 , further comprising:
building, by the one or more processors, a plurality of external characteristics associated with an external item by harvesting information about the external item from a website;
creating, by the one or more processors, an external vector that represents the external item based on the plurality of external characteristics using a deep learning model;
adding, by the one or more processors, the external vector to the plurality of learned vectors.

8. A system, comprising:
a memory, and
at least one processor coupled to the memory and configured to:
estimate a dimensional weight for an item purchased by a first user in an online market, wherein the item is shipped to a destination associated with the first user from an origin associated with a second user using a carrier, wherein the item is associated with a plurality of characteristics, and wherein the dimensional weight is determined by referencing a plurality of learned vectors each representing a previously shipped item;
create a vector that represents the item based on the plurality of characteristics;
determine one or more nearest vectors among the plurality of learned vectors by calculating distances between the vector that represents the item and the plurality of learned vectors to determine the one or more nearest vectors, wherein each vector in the one or more nearest vectors comprises an actual dimensional weight;
determine the dimensional weight of the item purchased by the first user, by calculating a weighted average of the actual dimensional weight for each vector in the one or more nearest vectors, wherein the actual dimensional weight for a first set of the one or more nearest vectors is adjusted up for heavier weights;
determine an estimated cost based on the dimensional weight of the item and the carrier; and
schedule a delivery of the item with the carrier for the estimated cost.

9. The system of claim 8 , wherein the calculating the weighted average employs a cost function that biases the weighted average towards a heavier estimate.

10. The system of claim 8 , wherein the carrier is a local carrier.

11. The system of claim 8 , wherein to determine the estimated cost, the at least one processor is further configured to:
send a query comprising the dimensional weight, the origin, and the destination to an application programming interface offered by the carrier and receive the estimated cost in response.

12. The system of claim 8 , wherein to schedule the delivery, the at least one processor is further configured to:
send the estimated cost to the second user;
receive a confirmation of the estimated cost, the origin, and a pickup time from the second user;
receive a confirmation of the destination and a drop-off time from the first user;
access an application programming interface offered by the carrier to place a shipping order comprising the item, the dimensional weight, the origin, the pickup time, the destination, and the drop-off time.

13. A non-transitory computer-readable device having instructions stored thereon that, when executed by at least one computing device, cause the at least one computing device to perform operations comprising:
estimating a dimensional weight for an item purchased by a first user in an online market, wherein the item is shipped to a destination associated with the first user from an origin associated with a second user using a carrier, wherein the item is associated with a plurality of characteristics, and wherein the dimensional weight is determined by referencing a plurality of learned vectors each representing a previously shipped item;
creating a vector that represents the item based on the plurality of characteristics;
determining one or more nearest vectors among the plurality of learned vectors by calculating distances between the vector that represents the item and the plurality of learned vectors to determine the one or more nearest vectors, wherein each vector in the one or more nearest vectors comprises an actual dimensional weight;
determining the dimensional weight of the item purchased by the first user, by calculating a weighted average of the actual dimensional weight for each vector in the one or more nearest vectors, wherein the actual dimensional weight for a first set of the one or more nearest vectors is adjusted up for heavier weights;
determining an estimated cost based on the dimensional weight of the item and the carrier; and
scheduling a delivery of the item with the carrier for the estimated cost.

14. The non-transitory computer-readable device of claim 13 , wherein the carrier is a local carrier.

15. The non-transitory computer-readable device of claim 13 , wherei

Assignees

• Mercari Inc

Inventors

• Mozzami Mohammad-Mahdi
• Crate Lawrie
• Uehara Masanori
• Turner Eric
• Clark Robin
• Ishikawa Yu
• Osaka Chiaki

Classifications

• G06N3/088

  Non-supervised learning, e.g. competitive learning · CPC title

• G06F16/9537

  Spatial or temporal dependent retrieval, e.g. spatiotemporal queries · CPC title

• G06N3/08

  Learning methods · CPC title

• G06Q10/08345 Primary

  Pricing · CPC title

• G06Q10/0835

  Relationships between shipper or supplier and carriers · CPC title