User Attribute Prediction Using App Usage and Sleep Patterns
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing user attribute prediction methods, particularly for demographic information like age and gender, suffer from low accuracy, lack of universality, and robustness, and pose privacy concerns due to reliance on image and voice data.
Innovation Solution
A deep learning-based method that predicts user attributes using APP usage patterns and sleep patterns without requiring image or voice data, employing a deep neural network to extract correlations and using latent variables to learn weights, enhancing prediction accuracy and universality.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If image and voice data are collected for user attribute prediction, then prediction accuracy may be improved, but user privacy is compromised and data capture complexity increases
Solution Approach 1:
The patent extracts and removes the harmful elements (image and voice data collection) from the prediction system, replacing them with text-based demographic information and device usage data. This extraction eliminates privacy concerns while maintaining prediction functionality through alternative data sources that do not compromise user confidentiality.
Solution Approach 2:
The patent introduces text-based demographic information and device usage patterns as intermediary data sources that mediate between the need for accurate prediction and user privacy protection. These intermediaries provide sufficient predictive power without requiring direct collection of sensitive biometric data, thus resolving the contradiction between accuracy and privacy.
2Measurement precision
If multiple data sources including image and voice are used for prediction, then prediction accuracy improves, but device complexity and data processing requirements increase
Solution Approach 1:
The patent extracts and removes the complex data capture and processing requirements for image and voice data, retaining only text-based demographic information and simple device usage metrics. This extraction significantly reduces device complexity and data processing requirements while preserving the core prediction functionality.
3Measurement precision
If extensive user data is collected to improve prediction accuracy, then more comprehensive user profiles can be constructed, but network throughput and data storage requirements increase
Solution Approach 1:
The patent extracts and removes the requirement for extensive data collection and transmission, relying instead on compact text-based demographic information and minimal device usage data. This extraction dramatically reduces network throughput consumption and data storage requirements while maintaining sufficient prediction accuracy for constructing user profiles.
Data Source
Figure 1~3
Figure 4
Figure 5
AI summary
The embodiment of the present disclosure provides a method, an apparatus, an electronic device, and a storage medium for predicting user attribute. The method comprises acquiring an input of demographic-related information; determining features corresponding to the input of the demographic-related information and a weight corresponding to each feature, where the features comprise a single feature corresponding to each input of the demographic-related information, and/or a cross feature between the at least two single features; predicting user attribute according to each feature and its corresponding weight. Based on the solution provided by the embodiment of the present disclosure, the accuracy on the prediction of the user attribute can be effectively improved.