Advertisement click rate estimation method based on bilinear FFM and multi-head attention mechanism
An advertising click and attention technology, applied in computer components, character and pattern recognition, instruments, etc., can solve the problems of poor interpretability of high-level features, average effect, and large number of parameters, so as to achieve improved effect and interpretability Effect
Pending Publication Date: 2021-12-14
GUANGDONG UNIV OF TECH
View PDF0 Cites 0 Cited by
- Summary
- Abstract
- Description
- Claims
- Application Information
AI Technical Summary
Problems solved by technology
FM model training speed is fast, but the effect is not as good as FFM
However, FFM introduces domains on the basis of FM, which makes the constant number of FFM increase. This parameter is obviously unacceptable in the case of large-scale data online deployment. Therefore, although FFM has a good effect, due to the large number of parameters, Difficult to deploy and implement online
DNN is used in the high-order feature combination part, but DNN uses the product method to implicitly extract feature combination for high-order feature combination, resulting in poor interpretability of the combined high-order features and general effect.
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 moreImage
Smart Image Click on the blue labels to locate them in the text.
Smart ImageViewing Examples
Examples
Experimental program
Comparison scheme
Effect test
Embodiment 1
[0054] Such as figure 1 Said, based on the bilinear FFM and the multi-head attention mechanism, the advertising click-through rate estimation method includes the following steps:
[0055] S1: Obtain the click data set of the advertiser's item advertisement and perform data preprocessing on the numerical features and categorical features in the data set;
[0056] S2: Map categorical feature data and numerical features into Embedding vectors;
[0057] S3: Input the Embedding vector into the bilinear FFM structure and the multi-head attention mechanism structure respectively, and combine the low-order features and high-order features respectively to obtain the combination of low-order features and high-order features;
[0058] S4: Splicing the low-order feature combination and the high-order feature combination, and calculating according to the Sigmoid function, to obtain the feature combination data and the corresponding probability;
[0059] S5: Group the feature combination ...
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
Login to View More Abstract
The invention provides an advertisement click rate estimation method based on a bilinear FFM and a multi-head attention mechanism. The method comprises the following steps: obtaining an advertiser article advertisement click data set, preprocessing the data set to obtain category type feature data, and converting the category type feature data into Embedding vectors; processing the Embedding vectors into a low-order feature combination and a high-order feature combination through a bilinear FFM structure and a multi-head attention mechanism structure; splicing and calculating the low-order feature combination and the high-order feature combination to obtain feature combination data and corresponding probabilities; and performing descending sorting on the feature combination data according to the probabilities, and taking out an advertisement on top as a recommendation result of advertisement click rate estimation. According to the method, the bilinear FFM structure is adopted, so that a small number of parameters are added on the basis of an FM model to achieve the effect close to an FFM; and meanwhile, relevance between combined feature vectors is calculated through the multi-head attention mechanism structure, so that the effect and interpretability of advertisement click rate estimation are improved.
Description
technical field [0001] The invention relates to the technical field of data mining, in particular to a method for estimating the click-through rate of an advertisement based on a bilinear FFM and a multi-head attention mechanism. Background technique [0002] With the development of e-commerce, Internet advertisement has become a new media advertisement entering people's life. Usually, before placing an advertisement, an advertiser wishes to know the click-through rate of an advertisement placed on a certain advertising space on the website, and makes a decision to reserve an advertising space based on the known click-through rate. In order to provide advertisers with a basis for making decisions about pre-scheduled advertising positions, the click-through rate of advertisements placed on a certain advertising position can be estimated for reference by advertisers. The conventional method for estimating the advertisement click-through rate in the prior art is: to train the ...
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
Login to View More IPC IPC(8): G06Q30/02G06K9/62G06N20/20
CPCG06Q30/0242G06N20/20G06F18/22
Inventor 李卫军廖永
Owner GUANGDONG UNIV OF TECH