Method and device for processing user interaction information

Patent metadata

Field	Value
Publication number	US-12062060-B2
Application number	US-202017604283-A
Country	US
Kind code	B2
Filing date	Mar 26, 2020
Priority date	Apr 15, 2019
Publication date	Aug 13, 2024
Grant date	Aug 13, 2024

Abstract

Embodiments of the present disclosure disclose a method and a device for processing user interaction information. A specific embodiment of the method comprises: acquiring a set of user interaction information associated with a preset interaction operation, wherein the user interaction information comprises category information and brand information of an interaction object, user attribute information, and operation time information of interaction operations corresponding to a brand of the interaction object; generating a corresponding interaction feature of a user on the basis of the set of user interaction information; and determining, on the basis of the interaction feature of the user and a pre-trained preset operation probability generation model, a probability of the user executing a target operation associated with a brand of the interaction object in the corresponding user interaction information.

First claim

What is claimed is:

1. A method for processing user interaction information, comprising:
acquiring a set of user interaction information associated with a preset interaction operation, wherein the user interaction information comprises: category information and brand information of an interaction object, user attribute information, and operation time information of an interaction operation corresponding to a brand of the interaction object;
generating, based on the set of user interaction information, a corresponding user interaction feature;
determining, based on the user interaction feature and a pre-trained preset operation probability generation model, probabilities of users performing a target operation associated with a target brand of a target interaction object in corresponding user interaction information, wherein the target operation comprises purchasing a commodity of the target brand;
determining probabilities greater than a preset threshold from the probabilities of the users performing the target operation associated with the target brand, and determining a number of the probabilities greater than the preset threshold;
generating inventory adjustment information according to the number of the probabilities greater than the preset threshold; and
controlling a target device to distribute commodities in a logistics warehouse based on the inventory adjustment information.

2. The method according to claim 1 , wherein the pre-trained preset operation probability generation model comprises a long short-term memory network, a first fully-connected network, a second fully-connected network, and a third fully-connected network.

3. The method according to claim 2 , wherein the user interaction information further comprises a display position of the interaction object, and the user interaction feature comprises an interaction operation feature matrix, a user attribute feature vector, a category feature vector, and a brand feature vector; and
generating, based on the set of user interaction information, the corresponding user interaction feature comprises:
generating, based on the user interaction information, a corresponding initial user interaction operation feature matrix, wherein an element in the initial interaction operation feature matrix is used to represent an interaction operation feature corresponding to the brand of the interaction object, and a row number and a column number of the element in the initial interaction operation feature matrix are used to identify an operation time of the interaction operation corresponding to the brand of the interaction object and the display position of the interaction object respectively;
converting the initial user interaction operation feature matrix into a corresponding two-dimensional matrix, and using the two-dimensional matrix as a corresponding user interaction operation feature matrix;
acquiring the user attribute feature vector generated based on the user attribute information in the user interaction information; and
acquiring the category feature vector generated based on information associated with a category of the interaction object in the user interaction information and the brand feature vector generated based on information associated with the brand of the interaction object in the user interaction information.

4. The method according to claim 3 , wherein the pre-trained preset operation probability generation model comprises at least one preset operation probability generation sub-model corresponding to the category; and
determining, based on the user interaction feature and the pre-trained preset operation probability generation model, the probability of the user performing the target operation associated with the target brand of target the interaction object in the corresponding user interaction information comprises:
inputting the user interaction feature generated according to the set of user interaction information into the preset operation probability generation sub-model matched with the category of the interaction object corresponding to the input interaction feature, to generate the probability, corresponding to the input interaction feature, of the user performing the target operation associated with the target brand of the target interaction object in the user interaction information.

5. The method according to claim 4 , wherein inputting the user interaction feature generated according to the set of user interaction information into the preset operation probability generation sub-model matched with the category of the interaction object corresponding to the input interaction feature, to generate the probability, corresponding to the input interaction feature, of the user performing the target operation associated with the target brand of the target interaction object in the user interaction information, comprises:
inputting the user interaction operation feature matrix generated according to the set of user interaction information into the long short-term memory network in the preset operation probability generation sub-model matched with the category corresponding to the input user interaction feature, to generate a corresponding first implicit feature;
inputting the user attribute feature vector generated according to the set of user interaction information into the first fully-connected network in the preset operation probability generation sub-model matched with the category corresponding to the input user interaction feature, to generate a corresponding second implicit feature;
inputting the category feature vector and the brand feature vector generated according to the set of user interaction information into the second fully-connected network in the preset operation probability generation sub-model matched with the category corresponding to the input user interaction feature, to generate a corresponding third implicit feature; and
inputting the generated first implicit feature, the second implicit feature, and the third implicit feature into the third fully-connected network in the preset operation probability generation sub-model matched with the category corresponding to the input user interaction feature, to generate the probability, corresponding to the input user interaction feature, of the user performing the target operation associated with the target brand of the target interaction object in the user interaction information.

6. The method according to claim 5 , wherein the preset operation probability generation sub-model is generated by training through following steps of:
acquiring a set of training samples, wherein a training sample comprises sample user interaction information and sample labeling information corresponding to the sample user interaction information, the sample user interaction information comprises sample category information and sample brand information of a sample interaction object, sample user attribute information and sample operation time information of a sample interaction operation corresponding to a sample brand of the sample interaction object, the sample labeling information is used to represent whether a sample user performs a sample target operation associated with a target sample brand of a target sample interaction object in corresponding sample user interaction information, and a category in each piece of sample user interaction information in the set of training samples is consistent;
generating, based on the sample user interaction information in the set of training samples, a corresponding sample user interaction feature; and
using the generated sample user interaction feature as an input, and using sample labeling information corresponding to the input sample interaction feature as an expected output, for training to obtain the preset operation probability ge

Assignees

• Beijing Wodong Tianjun Information Technology Co Ltd

Inventors

• Zhong Yu
• Du Ruihuan
• Cui Bo
• Shou Ruyang
• Lin Zhangang
• Chen Xi
• Nair Harikesh Sasikumar

Classifications

• G06N3/0442

  characterised by memory or gating, e.g. long short-term memory [LSTM] or gated recurrent units [GRU] · CPC title

• G06N3/09

  Supervised learning · CPC title

• G06N7/01

  Probabilistic graphical models, e.g. probabilistic networks · CPC title

• G06Q30/0201 Primary

  Market modelling; Market analysis; Collecting market data · CPC title

• G06N3/045

  Combinations of networks · CPC title