User portrait inference method and device based on spatio-temporal movement data representation learning

Abstract

An embodiment of the invention provides a user portrait inference method and device based on spatio-temporal movement data representation learning. The method comprises the following steps: acquiringa plurality of users and place data accessed by the users, representing the similarity of a user space-time mode by side length weights of the users, representing the functional similarity of the places by side length weights of the places, and representing the frequency of the places accessed by the users by side length weights of the users and the places to obtain a mobile network with reservedsemantics; obtaining a user representation vector when the preset target function is minimum, inputting the user representation vector into a preset machine learning classification model, and obtaining an inference result of the user portrait; wherein the target function is constructed according to the three types of side length weights, the user representation vector and the place representationvector. According to the method, a large amount of feature generation and feature screening do not need to be carried out manually, the model training efficiency is high, the labor cost is effectivelysaved, the model performance can be effectively guaranteed, and then accurate user attribute inference based on mobile data is achieved.

Description

technical field [0001] The present invention relates to the field of user portrait inference, in particular to a method and device for inferring user portraits based on spatio-temporal movement data representation learning. Background technique [0002] User Demographic Inference refers to the task of inferring one or several specific demographic attributes (such as age, gender, education level, etc.) of users based on user characteristic data. With the widespread popularity of smart mobile devices, spatio-temporal mobile data has become the most widely collected user characteristics, which has great potential in user attribute inference tasks. Typical application scenarios of user inference based on spatio-temporal movement information can help application developers provide better personalized services and designs for their user groups in the development of personalized applications, such as personalized product recommendations and travel navigation services on e-commerce ...

AI Technical Summary

Problems solved by technology

However, such methods can only identify users who are close in time and space (such as neighbors who often meet), and cannot identify user groups that are far apart in time and space but have the same attributes (such as teachers working in different schools). Unable to achieve accurate user attribute inference

At the same time, in mature machine learning algorithms such as GBDT and XgBoost, spatio-temporal movement data cannot be directly used as model input due to its unstructured characteristics, and a large number of feature generation and feature screening are required manually. The training model is inefficient and labor-intensive. Expensive and cannot effectively guarantee model performance

Images