A method and device for delivering services to crowds

Abstract

The embodiment of this specification discloses a method and device for delivering services to groups of people based on a prediction model. The method includes: obtaining a plurality of first groups of people, and obtaining group characteristics corresponding to each first group of people; inputting the group characteristics respectively a predictive model, to obtain the respective scores of the plurality of first groups of people from the output of the model, and the scores predict the weighted sum of the click rate and usage rate of the service by the first group of people; and based on the The scores of each of the multiple first groups of people are used to determine the group to deliver the service to.

Description

technical field [0001] The embodiments of this specification relate to the field of machine learning technology, and more specifically, to a method and device for delivering services to people based on a prediction model, and a method and device for training the prediction model. Background technique [0002] In the Internet, it is a very common marketing method to conduct business marketing to the crowd based on Internet data. In marketing activities, a large number of labels can usually be generated through qualitative research methods, and different groups of people can be established according to the labels. For example, by studying the user life cycle, we divide the process from potential customer to user to churn into 7 stages. If there are 3-10 significant features in each stage, then the entire user life cycle may have more than 3^7 feature combinations As a result, in actual marketing activities, the results of qualitative research cannot be directly applied to the...

AI Technical Summary

Problems solved by technology

For example, by studying the user life cycle, we divide the process from potential customer to user to churn into 7 stages. If there are 3-10 significant features in each stage, then the entire user life cycle may have more than 3^7 feature combinations As a result, in actual marketing activities, the results of qualitative research cannot be directly applied to the circle of people, and only some people can be found for marketing through manual sampling.

In addition, especially for newly established business platforms, due to the lack of sufficient data accumulation, traditional marketing schemes such as tag circle people and algorithm circle people cannot obtain effective marketing effects

Images