Information flow media advertisement creative recommendation method and device

Abstract

The invention discloses an information flow media advertisement creative recommendation method and device, and the method comprises the steps: constructing bottom storage, and storing advertisement putting data in a database corresponding to the bottom storage; and acquiring the copywriting data, the material data, the creative effect data and the user group data from the database; performing datapreprocessing on the copywriting data, the material data, the creative effect data and the user crowd data; performing data mining and feature engineering on the preprocessed copywriting data, material data, creative effect data and user population data to obtain a user vector, a creative vector and an One-Hot vector; and inputting the user vector, the creativity vector and the One-Hot vector into a Wide & Deep model, and outputting the probability y of creativity recommendation to the corresponding user crowd by the Wide & Deep model, y belonging to [0, 1]; and performing effect evaluation according to the probability of creatively recommending to the corresponding user crowd. According to the information flow media advertisement creative recommendation method and device, the click rateof a user during creativity watching can be increased.

Description

technical field [0001] The present invention relates to the technical field of advertisement recommendation, in particular to a method and device for creative recommendation of information streaming media advertisements. Background technique [0002] Creativity is composed of materials (pictures) and copywriting (text). Currently, in the field of information streaming media creative recommendation, the mainstream structure adopted is to input materials, copywriting content and creative effect data, and use machine learning models to construct CTR prediction models. CTR (Click-Through-Rate) is the click-through rate, which specifically refers to the actual number of clicks on the idea divided by the amount of display of the idea. [0003] The traditional CTR estimation solution in the industry is the generalized linear model LR (logistic regression, logistic regression) + artificial feature engineering. LR uses the Logit transformation to map the function value to the [0,1] ...

AI Technical Summary

Problems solved by technology

However, the disadvantage of this solution is that because the learning ability of the linear model is limited, it is necessary to introduce a large amount of domain knowledge to manually design features and cross combinations between features to indirectly supplement the nonlinear learning ability of the algorithm, which consumes a lot of manpower and machine resources. Migration is not friendly enough

In addition, there is a common Tree based method. Although this method can effectively solve the feature combination problem of the LR model, the disadvantage is that it still remembers historical behavior and has poor generalization ability.

There are also FM (factorization machine) models, which can automatically learn the weights of high-order attributes without manually selecting features for crossover, but the FM model can only fit specific nonlinear models and cannot automatically perform feature crossover. The amount grows exponentially with the amount of data

Moreover, since the current mainstream models generally only use the characteristic variables of creativity (materials and copywriting) and its effect data, and do not establish contact with the recommended user groups, the offline effect of the creative recommendation model is better, but the online performance is not satisfactory.

Images