Binary classification model training method and device and multimedia data classification method and device

Abstract

The invention discloses a binary classification model training method and device and a multimedia data classification method and device. According to the invention, the method comprises the steps: after a business sample data set is obtained, acquiring the occurrence probability of each piece of sample data in the business sample data set, acquiring the probability distribution corresponding to the business sample data set and the distribution characteristics of the probability distribution, and selecting a to-be-learned posterior distribution function corresponding to the distribution characteristics of the probability distribution; based on a preset sampling algorithm and the to-be-learned parameter, sampling sample data in the probability distribution, and obtaining a target parameter value of the to-be-learned parameter and a posterior distribution function carrying the target parameter value so as to obtain a target threshold satisfying the posterior distribution function; performing positive and negative sample data division on each piece of sample data by adopting a target threshold value; obtaining a binary classification model based on the business data based on the divided positive sample data, the negative sample data and the feature information of the business object corresponding to each sample data. According to the method, the classification accuracy of the dichotomy model is improved.

Description

technical field [0001] The present disclosure relates to the technical field of the Internet, in particular to a method and device for training a binary classification model and classifying multimedia data. Background technique [0002] In business scenarios in the Internet field, we often encounter the problem of modeling and sorting business indicators such as video playback time, user online time, and the number of daily active users (Daily Active User, DAU). These indicators are usually continuous variables. For example, when users are divided into high-active users and low-active users based on the business indicator of video playback time, it is necessary to set an appropriate threshold for video playback time, and determine positive sample labels, negative sample labels, and the behavior of each account based on the threshold. The data trains the binary classification model to perform binary classification on the video playback time corresponding to each user to obtai...

AI Technical Summary

Problems solved by technology

[0007] However, in method 1, the artificial threshold value only depends on the user's understanding of the business, which is highly subjective, resulting in low classification accuracy; in method 2, the threshold value is set according to the statistics of the sample data, which is vulnerable to abnormalities in the sample data. Influenced by the value, the accuracy of classification is not high; in method 3, when the fluctuation of the variable value of the continuous variable is relatively large, the effect of regression fitting is average, and the effect of regression prediction cannot measure the accuracy of the final sorting of samples

[0008] In other words, the classification accuracy of the binary classification model using the above classification method is not high

Images