US 11,386,353 B2Granted

Method and apparatus for training classification model, and method and apparatus for classifying data

AssigneeTencent Technology (Shenzhen) Company Limited

InventorHongjun Yin

Life of this patent

Estimating the term…

Loading patent…

Abstract

This application discloses a method and an apparatus for training a classification model. The method includes obtaining a training sample, the training sample including a training parameter and a true classification corresponding to the training parameter and preforming classification training on an initial classification model by using the training parameter, to obtain a predicted classification. The method also includes determining a residual between the true classification and the predicted classification according to a gradient loss function of the initial classification model, the gradient loss function comprising a distance factor representing a distance between a first category and a second category, the first category being a category to which the predicted classification belongs, and the second category being a category to which the true classification belongs. The method further includes modifying the initial classification model according to the residual to obtain a final classification model.

Claims20

Dependent claims

A method for training a classification model for education level classification, the method comprising:

obtaining, by a device comprising a memory storing instructions and a processor in communication with the memory, a training sample, the training sample comprising a training parameter and a true classification corresponding to the training parameter, wherein the training parameter comprises the following parameters:
a type of reading;
a type of joined activity;
and
a subscribed official account;
performing, by the device, classification training on an initial classification model by using the training parameter, to obtain a predicted classification;
determining, by the device, a residual between the true classification and the predicted classification according to a gradient loss function of the initial classification model, the gradient loss function being based on a modification prediction function which is modified from a prediction function according to a distance vector, wherein:
the number of elements in the distance vector equals to the number of education level classifications;
one of the elements in the distance vector represents a distance factor;
the distance factor represents a distance between a first category and a second category;
rest of the elements in the distance vector are same constant numbers;
and
the first category and the second category each comprises the following categories:
a doctoral degree category;
a master's degree category;
a bachelor degree category;
a college degree category;
a high school degree category;
a junior high school degree category;
and
an elementary school degree category;
the doctoral degree category, the master's degree category, the bachelor degree category, the college degree category, the high school degree category, the junior high school degree category, and the elementary school degree category correspond to label values following a sequential order;
and
the distance factor is based on a difference between a label value of the first category and a label value of the second category;
modifying, by the device, the initial classification model according to the residual to obtain a final classification model.

22 more elements

An apparatus for training a classification model for education level classification, comprising:

a memory storing instructions;
and
a processor in communication with the memory, wherein, when the processor executes the instructions, the processor is configured to cause the apparatus to:
obtain a training sample, the training sample comprising a training parameter and a true classification corresponding to the training parameter, wherein the training parameter comprises the following parameters:
a type of reading;
a type of joined activity;
and
a subscribed official account;
perform classification training on an initial classification model by using the training parameter, to obtain a predicted classification;
determine, according to a gradient loss function of the initial classification model, a residual between the true classification and the predicted classification, the gradient loss function being based on a modification prediction function which is modified from a prediction function according to a distance vector, wherein:
the number of elements in the distance vector equals to the number of education level classifications;
one of the elements in the distance vector represents a distance factor;
the distance factor represents a distance between a first category and a second category;
rest of the elements in the distance vector are same constant numbers;
and
the first category and the second category each comprises the following categories:
a doctoral degree category;
a master's degree category;
p4 a bachelor degree category;
a college degree category;
a high school degree category;
a junior high school degree category;
and
an elementary school degree category;
the doctoral degree category, the master's degree category, the bachelor degree category, the college degree category, the high school degree category, the junior high school degree category, and the elementary school degree category correspond to label values following a sequential order;
and
the distance factor is based on a difference between a label value of the first category and a label value of the second category;
and
modify, according to the residual, the initial classification model to obtain a final classification model.

26 more elements

A non-transitory computer readable storage medium storing instructions, wherein the instructions, when executed by a processor, cause the processor to:

obtain a training sample, the training sample comprising a training parameter and a true classification corresponding to the training parameter, wherein the training parameter comprises the following parameters:
a type of reading;
a type of joined activity;
and
a subscribed official account;
perform classification training on an initial classification model by using the training parameter, to obtain a predicted classification;
determine, according to a gradient loss function of the initial classification model, a residual between the true classification and the predicted classification, the gradient loss function being based on a modification prediction function which is modified from a prediction function according to a distance vector, wherein:
the number of elements in the distance vector equals to the number of education level classifications;
one of the elements in the distance vector represents a distance factor;
the distance factor represents a distance between a first category and a second category;
rest of the elements in the distance vector are same constant numbers;
and
the first category and the second category each comprises the following categories:
a doctoral degree category;
a master's degree category;
a bachelor degree category;
a college degree category;
a high school degree category;
a junior high school degree category;
and
an elementary school degree category;
the doctoral degree category, the master's degree category, the bachelor degree category, the college degree category, the high school degree category, the junior high school degree category, and the elementary school degree category correspond to label values following a sequential order;
and
the distance factor is based on a difference between a label value of the first category and a label value of the second category;
and
modify, according to the residual, the initial classification model to obtain a final classification model.

23 more elements

Dependencies are inferred from the claim text.

Facts

Application: 16/286,894Filed 27 Feb 2019 · continuation
Priority: CN 201611139498.512 Dec 2016
Published: US 2019/0197429 A127 Jun 2019
Granted: 12 Jul 2022
Term ends: Not estimatedThe source records don’t support an estimate.

Classification4 codes

G06N 20/20
Machine learning: ensemble learning
G06N 5/01
Computing arrangements using knowledge-based models: dynamic search techniques; Heuristics; Dynamic trees; Branch-and-bound
G06N 20/00
Machine learning
G06N
AI and machine learning