Marketing prediction method combining automatic feature engineering and residual neural network

Abstract

The invention discloses a marketing prediction method combining automatic feature engineering and a residual neural network. The marketing prediction method comprises a data preprocessing step, a dataset division step, a model establishment step, a model prediction step and a model evaluation and optimization step. Compared with traditional collaborative filtering adopted in the field of data marketing using operator data, the method effectively uses the characteristics of an embedded layer in a neural network to perform effective information extraction and dimensionality reduction on high-dimensional sparse characteristics formed after class characteristics in the CTR problem are numbered in a dumb mode; and a residual network is constructed through short-circuit connection, so that theproblem of gradient disappearance in the training process is effectively solved, an approach for directly predicting the advertisement click willingness of a user can be provided, and the method is also suitable for processing data with large-scale sparse characteristics in recommendation systems such as advertisement marketing and the like.

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 combined with automatic feature engineering and residual neural network. Background technique [0002] With the rapid development of economic globalization and market economy, advertising marketing activities play an increasingly important role in the marketing strategy of enterprises, and are an important part of the marketing mix of enterprises. Online advertising marketing is to maximize the spread to the audience with the help of online marketing, and it is also more accurate. Online advertising marketing requires advertisers to use the network platform to put advertisements to target customers. [0003] The big data intelligent customer acquisition system, centered on the operator's large database, directly captures the contact information of users who meet the custom conditions, and...

AI Technical Summary

Problems solved by technology

[0007] ①. Because the linear model is relatively simple and has limited expressive ability, it has great limitations in the interaction between features and the construction of high-order features.

[0009] ②. Those skilled in the art know that in deep learning, the feature information in the sparse matrix can be concentrated and extracted through the added embedding layer, and the dimensionality reduction can be effectively realized; however, it is found in many practices that In the process of error backpropagation, the derivation of the loss function can easily cause the problem of gradient disappearance

Images