Marketing prediction method combining inner/outer product feature interaction and Bayesian neural network

Abstract

The invention discloses a marketing prediction method combining inner / outer product feature interaction and a Bayesian neural network. The method comprises a data preprocessing step, a data set division step, a model establishment step and a marketing activity click prediction step. According to the method, in the building process of a prediction model, the prediction uncertainty is introduced into the Bayesian neural network by effectively utilizing Bayesian inference, so that the Bayesian neural network model has higher robustness. And by adopting an inner / outer product combination method, the features are crossed to extract high-dimensional recessive features. Therefore, the application of deep learning to advertisement calculation and recommendation system algorithm problems can be effectively expanded, and the accuracy of user click behavior prediction is remarkably improved.

Description

technical field [0001] The present invention relates to the technical field of artificial intelligence in Internet marketing, and more specifically, relates to a marketing prediction method combining inner / outer product feature interaction and Bayesian neural network. Background technique [0002] Online advertising marketing is to maximize the spread to the audience with the help of online marketing, and it uses the network platform to put advertisements to target customers. In computing advertising and recommendation system algorithms, commonly used algorithms include linear models such as logistic regression (logistic regression, LR), factorization machine (factorization machine, FM) and so on. [0003] The above-mentioned algorithms have the characteristics of good explainability and simple algorithm implementation. However, due to the simplicity of the algorithm itself, the expressive ability is limited. Therefore, these algorithms are often difficult to extract high-o...

AI Technical Summary

Problems solved by technology

[0006] ①. In most recommendation system data sets, a large-dimensional sparse matrix will be formed, that is, a matrix composed of 0 and 1. For deep learning models based on gradient descent, there are certain difficulties; at the same time, large sparse matrices are also Will cause large computing power consumption and excessive computing time

[0008] ③. The traditional deep learning model directly completes the intersection and combination of features through multi-layer fully connected layers, but this method lacks certain "pertinence"

[0009] First of all, the fully connected layer does not cross between different feature domains;

Images