Predicting a status of a transaction

• US 10,217,148 B2
• Filed: 05/29/2015
• Issued: 02/26/2019
• Est. Priority Date: 01/23/2015
• Status: Active Grant

• Current Assignee: eBay Inc.
• Original Assignee: eBay Inc.
• Inventors: Abbasi Moghaddam, Samaneh, Louis, Jerry, Dalal, Vipul C.
• Primary Examiner(s): Frenel, Vanel
• Application Number:
  US14/725,977
• Publication Number:
  US 20160217513A1
• Time in Patent Office: 1,369 Days
• Field of Search: 705 2635, 705 261, 705301, 705348, 705 712, 705 738, 709248, 707758
• CPC Class Codes: G06Q 10/067 Enterprise or organisation modelling
• CPC Class Codes: G06Q 30/0609 Buyer or seller confidence or verification

Abstract

Methods, systems, and apparatus for predicting a status of a transaction are described. Feature data related to one or more transactions is collected and a subset of features is selected for use in predicting the status of the transaction. A model is trained using the collected feature data that corresponds to the selected features, and the model is applied to feature data of a selected transaction to generate a probability of the selected transaction attaining one or more defined statuses. Mitigating or preventive actions are performed based on the generated probability.

Claims

1. A system for predicting a status of a transaction, the system comprising:

a data collection module implemented on a transaction prediction server, the data collection module collecting feature data related to one or more transactions;

a model generation module implemented on the transaction prediction server, the model generation module selecting a subset of features for use in predicting the status of the transaction and training a model using the collected feature data that corresponds to the selected subset of features;

a prediction processing module implemented on the transaction prediction server, the prediction processing module applying the model to feature data of a selected transaction between a buyer and a seller to generate a probability of the selected transaction attaining one or more defined statuses; and

a transaction mitigation module implemented on the transaction prediction server, the transaction mitigation module configured to perform one or more mitigating or preventive actions based on the generated probability, the one or more mitigating or preventive actions determined based on a user-defined mapping between probability rule conditions corresponding to the one or more defined statuses and corresponding actions to be performed, wherein the one or more mitigating or preventive actions comprise one or more of blocking the selected transaction, communicating a warning to the buyer and the seller via user devices corresponding to the buyer and seller, or holding funds due to the seller.

2. The system of claim 1, wherein the data collection module is further configured to normalize the collected feature data.

3. The system of claim 1, wherein the feature data is collected for one or more of a buyer, a seller, an item, and a transaction.

4. The system of claim 1, wherein the collected feature data comprises one or more of user segment data, user historical data, user behavioral data, an item price, an item title, an item category, a transaction site identifier, a buyer tool, a shipping fee, and a shipping time.

5. The system of claim 1, wherein the model is a Logistic Regression model.

6. The system of claim 5, wherein the subset of features is selected using a Forward Feature Selection mechanism of the Logistic Regression model.

7. The system of claim 1, wherein the model is a Random Forest model.

8. The system of claim 7, wherein the subset of features is selected using a Recursive Feature Elimination mechanism of the Random Forest model.

9. The system of claim 1, wherein the model is trained for each of a plurality of leaf-categories of items.

10. The system of claim 1, wherein the one or more mitigating or preventive actions are one or more of issuing a message to a buyer, issuing a message to a seller, blocking the buyer from executing the transaction, blocking the seller from executing the transaction, and holding funds of the seller.

11. A method for predicting a status of a transaction, the method comprising:

collecting feature data related to one or more transactions;

selecting a subset of features for use in predicting the status of the transaction;

training a model using the collected feature data that corresponds to the selected subset of features;

applying the model to feature data of a selected transaction to generate a probability of the selected transaction attaining one or more defined statuses;

determining one or more mitigating or preventive actions, the determining based on a user-defined mapping between probability rule conditions corresponding to the one or more defined statuses and corresponding actions to be performed; and

performing the one or more mitigating or preventive actions, the one or more mitigating or preventive actions being one or more of blocking the selected transaction, communicating a warning to the buyer and the seller via user devices corresponding to the buyer and seller, or holding funds due to the seller.

12. The method of claim 11, further comprising normalizing the collected feature data.

13. The method of claim 11, wherein the feature data is collected for one or more of a buyer, a seller, an item, and a transaction.

14. The method of claim 11, wherein the collected feature data comprises one or more of user segment data, user historical data, user behavioral data, an item price, an item title, an item category, a transaction site identifier, a buyer tool, a shipping fee, and a shipping time.

15. The method of claim 11, wherein the model is a Logistic Regression model.

16. The method of claim 15, wherein the subset of features is selected using a Forward Feature Selection mechanism of the Logistic Regression model.

17. The method of claim 11, wherein the model is a Random Forest model.

18. The method of claim 17, wherein the subset of features is selected using a Recursive Feature Elimination mechanism of the Random Forest model.

19. The method of claim 11, wherein the model is trained for each of a plurality of leaf-categories of items.

20. A non-transitory computer-readable medium embodying instructions that, when executed by a processor, perform operations comprising:

collecting feature data related to one or more transactions;

selecting a subset of features for use in predicting a status of at least one of the transactions;

training a model using the collected feature data that corresponds to the selected subset of features;

applying the model to feature data of a selected transaction to generate a probability of the selected transaction attaining one or more defined statuses;

determining one or more mitigating or preventive actions, the determining based on a user-defined mapping between probability rule conditions corresponding to the one or more defined statuses and corresponding actions to be performed; and

performing the one or more mitigating or preventive actions, the one or more mitigating or preventive actions being one or more of blocking the selected transaction, communicating a warning to the buyer and the seller via user devices corresponding to the buyer and seller, or holding funds due to the seller.