Method and system for training model by using training data

Patent metadata

Field	Value
Publication number	US-11615346-B2
Application number	US-201816112592-A
Country	US
Kind code	B2
Filing date	Aug 24, 2018
Priority date	Feb 25, 2016
Publication date	Mar 28, 2023
Grant date	Mar 28, 2023

Abstract

Embodiments of the present disclosure provide a method and system for training a model by using training data. The training data includes a plurality of samples, each sample includes N features, and features in the plurality of samples form N feature columns, and the method includes: determining an importance value of each of the N feature columns; determining whether the importance value of each of the N feature columns satisfies a threshold condition; performing a dimension reduction on M feature columns to generate P feature columns in response to the determination that the importance values of the M feature columns do not satisfy the threshold condition, wherein M<N and P<M; merging (N−M) feature columns having importance values that satisfy the threshold condition and the generated P feature columns to obtain (N−M+P) feature columns; and training the model based on the training data including the (N−M+P) feature columns.

First claim

What is claimed is:

1. A method for training a model by using training data, wherein the training data comprises a plurality of samples, each sample comprises N features, and features in the plurality of samples form N feature columns, and the method comprises:
determining an importance value of each of the N feature columns;
determining whether the importance value of each of the N feature columns satisfies a threshold condition;
performing, among the N feature columns, a dimension reduction on M feature columns to generate P feature columns in response to a determination that importance values of the M feature columns do not satisfy the threshold condition, wherein M<N and P<M;
merging (N−M) feature columns having importance values that satisfy the threshold condition and the generated P feature columns to obtain (N−M+P) feature columns; and
training the model based on the training data including the (N−M+P) feature columns.

2. The method according to claim 1 , wherein performing, among the N feature columns, the dimension reduction on M feature columns to generate P feature columns in response to the determination that the importance values of the M feature columns do not satisfy the threshold condition further comprises:
in response to the determination that the importance values of M feature columns in the N feature columns do not satisfy the threshold condition, performing dimension reduction processing on the M feature columns by using a minwise Hashing algorithm to generate P feature columns, wherein P=k×2 b , and k and b are specified by the minwise Hashing algorithm.

3. The method according to claim 1 , wherein the importance values of the (N−M) feature columns are greater than or equal to a threshold.

4. The method according to claim 1 , wherein the training data further comprises a tag value corresponding to each sample, and determining an importance value of each of the N feature columns comprises:
determining at least one of an information value and an information gain of each of the N feature columns by using the tag value and features in each of the N feature columns, and
using the at least one of the information value and the information gain as the importance value.

5. The method according to claim 4 , wherein a threshold corresponding to the information value is a first threshold, and a threshold corresponding to the information gain is a second threshold.

6. The method according to claim 1 , wherein the model is a classification algorithm model or a regression algorithm model.

7. A system for training a model by using training data, wherein the training data comprises a plurality of samples, each sample comprises N features, and features in the samples form N feature columns, and the system comprises:
a memory storing a set of instructions; and
one or more processors configured to execute the set of instructions to cause the system to perform:
determining an importance value of each of the N feature columns;
determining whether the importance value of each feature column satisfies a threshold condition;
performing, among the N feature columns, a dimension reduction on M feature columns to generate P feature columns in response to a determination that importance values of the M feature columns in the N feature columns do not satisfy the threshold condition, wherein M<N and P<M;
merging (N−M) feature columns having importance values that satisfy the threshold condition and the generated P feature columns to obtain (N−M+P) feature columns; and
training the model based on the training data comprising the (N−M+P) feature columns.

8. The system according to claim 7 , wherein the one or more processors are configured to execute the set of instructions to cause the system to further perform:
performing the dimension reduction on the M feature columns by using a minwise Hashing algorithm to generate the P feature columns, wherein P=k×2 b , and k and b are specified by the minwise Hashing algorithm.

9. The system according to claim 7 , wherein the training data further comprises a tag value corresponding to each sample, and the one or more processors are configured to execute the set of instructions to cause the system to further perform:
determining at least one of an information value and an information gain of each of the N feature columns by using the tag value and features in each of the N feature columns, and
using the at least one of the information value and the information gain as the importance value.

10. The system according to claim 9 , wherein a threshold corresponding to the information value is a first threshold, and a threshold corresponding to the information gain is a second threshold.

11. The system according to claim 7 , wherein the model is a classification algorithm model or a regression algorithm model.

12. The system according to claim 7 , wherein the importance values of the (N−M) feature columns are greater than or equal to a threshold.

13. A non-transitory computer readable medium that stores a set of instructions that is executable by at least one processor of a computer system to cause the computer system to perform a method for training a model by using training data, wherein the training data comprises a plurality of samples, each sample comprises N features, and features in the plurality of samples form N feature columns, and the method comprises:
determining an importance value of each of the N feature columns;
determining whether the importance value of each of the N feature columns satisfies a threshold condition;
performing, among the N feature columns, a dimension reduction on M feature columns to generate P feature columns in response to a determination that importance values of the M feature columns do not satisfy the threshold condition, wherein M<N and P<M;
merging (N−M) feature columns having importance values that satisfy the threshold condition and the generated P feature columns to obtain (N−M+P) feature columns; and
training the model based on the training data including the (N−M+P) feature columns.

14. The non-transitory computer readable medium according to claim 13 , wherein performing, among the N feature columns, the dimension reduction on M feature columns to generate P feature columns in response to the determination that the importance values of the M feature columns do not satisfy the threshold condition further comprises:
in response to the determination that the importance values of M feature columns in the N feature columns do not satisfy the threshold condition, performing dimension reduction processing on the M feature columns by using a minwise Hashing algorithm to generate P feature columns, wherein P=k×2 b , and k and b are specified by the minwise Hashing algorithm.

15. The non-transitory computer readable medium according to claim 13 , wherein the importance values of the (N-M) feature columns are greater than or equal to a threshold.

16. The non-transitory computer readable medium according to claim 13 , wherein the training data further comprises a tag value corresponding to each sample, and determining an importance value of each of the N feature columns comprises:
determining at least one of an information value and an information gain of each of the N feature columns by using the tag value and features in each of the N feature columns, and
using the at least one of the information value and the information gain as the importance value.

17. The non-transitory computer readable medium according to claim 16 , wherein a threshold corresponding to the information value is a

Assignees

• Alibaba Group Holding Ltd

Inventors

• Dai Bin
• Li Shen
• Jiang Xiaoyan
• Yang Xu
• Qi Yuan
• Chu Wei
• Wang Shaomeng
• Fu Zihao

Classifications

• G06N20/20 Primary

  Ensemble learning · CPC title

• G06F18/2137

  based on criteria of topology preservation, e.g. multidimensional scaling or self-organising maps · CPC title

• G06F18/217

  Validation; Performance evaluation; Active pattern learning techniques · CPC title

• G06N20/00 Primary

  Machine learning · CPC title

• G06F18/2148

  characterised by the process organisation or structure, e.g. boosting cascade · CPC title