Looking for breakthrough ideas for innovation challenges? Try Patsnap Eureka!

A Click-through Rate Prediction Method Based on Multi-task Learning Mechanism

A multi-task learning and click-through rate technology, applied in the field of recommendation systems, can solve problems such as limited model combination capabilities, achieve good scalability and improve effectiveness

Active Publication Date: 2022-05-10
ZHEJIANG UNIV
View PDF3 Cites 0 Cited by
  • Summary
  • Abstract
  • Description
  • Claims
  • Application Information

AI Technical Summary

Problems solved by technology

The combination of tree models has created a trend of automatic feature crossing, but the combination ability of the model is still very limited

Method used

the structure of the environmentally friendly knitted fabric provided by the present invention; figure 2 Flow chart of the yarn wrapping machine for environmentally friendly knitted fabrics and storage devices; image 3 Is the parameter map of the yarn covering machine
View more

Image

Smart Image Click on the blue labels to locate them in the text.
Viewing Examples
Smart Image
  • A Click-through Rate Prediction Method Based on Multi-task Learning Mechanism
  • A Click-through Rate Prediction Method Based on Multi-task Learning Mechanism
  • A Click-through Rate Prediction Method Based on Multi-task Learning Mechanism

Examples

Experimental program
Comparison scheme
Effect test

Embodiment

[0051] When predicting the degree of interest of user U in question Q, the processing flow of the system is described as follows:

[0052] 1) Use Scikit-learn and Numpy tools to perform feature representation on user U and question Q, expressed as one-hot encoded features in the following form:

[0053]

[0054] 2) All features are densely represented by embedding pooling. Meta-Expert performs knowledge extraction on question Q, and MoE performs multi-angle portraits on user I and I's browsing question set.

[0055] 3) The output of each Expert module and Meta-Expert module in the MoE calculates the gating signal, through the weighted pooling of the gating signal, and finally obtains the user interest representation of the corresponding task.

[0056] 4) The CTR and CVR tasks target user interest representation vectors, and use logistic regression to output corresponding prediction results. The value range is [0,1]. A larger value indicates a greater degree of interest. ...

the structure of the environmentally friendly knitted fabric provided by the present invention; figure 2 Flow chart of the yarn wrapping machine for environmentally friendly knitted fabrics and storage devices; image 3 Is the parameter map of the yarn covering machine
Login to View More

PUM

No PUM Login to View More

Abstract

The invention discloses a click-through-rate estimation (Click-Through-Rate, CTR) method based on a multi-task learning mechanism, aiming at improving the effect of personalized recommendation for users in a text content recommendation system. Based on the idea of ​​integrating domain knowledge into deep models to improve user interest mining, the present invention proposes a novel hybrid expert network to perform highly interpretable representation of domain knowledge, and models the interaction between multiple tasks through a task-oriented gating network. Contact to further improve the accuracy of CTR and other task predictions. The method proposed by the present invention can directly input user features and candidate item features in the form of one-hot encoding (One-Hot), without cumbersome manual feature engineering, and can learn the user's deep interest representation through the feature crossing of the depth model, Finally, a probability value in the range of 0 to 1 is output to indicate the user's interest in the candidate item. The present invention has high interpretability and expansibility, and can be easily applied to specific recommendation scenarios.

Description

technical field [0001] The present invention relates to CTR estimation, CVR estimation, multi-task learning mechanism, knowledge representation and other fields in the field of recommendation system, and specifically relates to a click rate estimation method based on multi-task learning mechanism. Background technique [0002] The recommendation system is to solve the problem of how to help users quickly filter redundant data and find the information they are interested in under the situation of "information overload". Current recommender systems play an important role in question answering communities. Recommend questions that may be of interest to users based on their preferences to improve user experience. However, the current recommendation algorithm research is facing technical difficulties such as the difficulty of mining users' implicit interests and how to effectively integrate domain knowledge. [0003] Take Zhihu and other open domain question-and-answer communit...

Claims

the structure of the environmentally friendly knitted fabric provided by the present invention; figure 2 Flow chart of the yarn wrapping machine for environmentally friendly knitted fabrics and storage devices; image 3 Is the parameter map of the yarn covering machine
Login to View More

Application Information

Patent Timeline
no application Login to View More
Patent Type & Authority Patents(China)
IPC IPC(8): G06F16/9535G06F16/2458G06F16/332G06Q50/00G06N3/04G06N3/08
CPCG06F16/9535G06F16/2465G06F16/3329G06Q50/01G06N3/08G06N3/048
Inventor 张引胡荐苛
Owner ZHEJIANG UNIV
Who we serve
  • R&D Engineer
  • R&D Manager
  • IP Professional
Why Patsnap Eureka
  • Industry Leading Data Capabilities
  • Powerful AI technology
  • Patent DNA Extraction
Social media
Patsnap Eureka Blog
Learn More
PatSnap group products

Browse by: Latest US Patents, China's latest patents, Technical Efficacy Thesaurus, Application Domain, Technology Topic, Popular Technical Reports.

© 2024 PatSnap. All rights reserved.Legal|Privacy policy|Modern Slavery Act Transparency Statement|Sitemap|About US| Contact US: help@patsnap.com