Click rate estimation model for feature interaction selection based on three-way decision theory

Abstract

A click rate estimation model for performing feature interaction selection based on a three-way decision theory is used for calculating a click rate estimation task in an advertisement and recommendation system, and the method comprises the following steps of: 1) processing input classification features by utilizing One-hot coding and an Embedding technology; 2) interacting the characteristics of the original data by using a factorization machine and a derivative model thereof; 3) selecting interaction features by using a three-way decision gate proposed by combining a three-way decision theory with a binary Sigmoid function; and 4) outputting the finally predicted click rate through a logistic regression function according to the interactive characteristics selected by the three decision gates and the original characteristics of the data. Noise information brought to an original model by excessive redundant interaction features can be reduced, important feature interaction is enhanced to a certain extent, general important interaction features are reserved, and the redundant interaction features are eliminated. The performance of the factorization machine and the derivative model thereof is improved, and meanwhile the training time of the model is shortened.

Description

technical field [0001] The invention belongs to the field of computational advertising and recommendation systems, and specifically relates to interactive features in click-through rate estimation, in particular to a click-through rate estimation model for feature interaction selection based on three-way decision-making theory. Background technique [0002] In recommender systems and online advertising, determining the probability of a user clicking on a specified product or advertisement is an important task, which determines the accuracy of the recommender system and the revenue of online advertising. In the Top-N recommendation of the recommendation system, the user's click rate on a product is an important basis for sorting. This click rate determines the position of the product in the recommendation list. The system often prefers to place products with a higher click rate in the recommendation top of the list to improve the accuracy of personalized recommendations. In ...

AI Technical Summary

Problems solved by technology

[0005] In order to solve the above problems, the present invention provides a click-through rate estimation model based on the three-way decision-making theory for feature interactive selection, which can not only achieve more accurate click-through rate estimation, but also meet the online and offline requirements in terms of model complexity. It provides a better click rate prediction model for online advertising platforms and e-commerce platforms

Images