Deep Learning Model Predicts User Demographics From Playlog Data
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
In the mobile app environment, companies face challenges in accessing user information related to advertisement identification information, such as ADID and IDFA, which limits their ability to perform customized marketing effectively.
Innovation Solution
A method and device using deep learning models to predict user information, such as gender and age, from playlog data, allowing for the generation of training data based on account identification information and subsequent prediction of user information related to advertisement identification information.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Productivity
If marketing is performed based on advertisement identification information using volume-based approaches, then the marketing can be implemented without access to user information, but the marketing efficiency and customization capability deteriorate
Solution Approach 1:
The patent introduces playlog data as an intermediary that bridges the gap between advertisement identification information and user information. The deep learning model uses playlog data (game play history, in-game purchases, social interactions) as intermediate features to infer user demographics, enabling customized marketing without direct access to user information tied to advertisement IDs
Solution Approach 2:
The patent replaces the traditional mechanical approach of direct user information access with a data-driven deep learning system. Instead of directly querying user databases using advertisement identification information, the system uses neural networks to predict user demographics from playlog patterns, substituting computational inference for direct information retrieval
2Adaptability or versatility
If user information such as gender and age is accessed to perform customized marketing, then marketing customization improves, but access to such information is restricted for advertisement identification information
Solution Approach 1:
Playlog data serves as an intermediary that indirectly reveals user demographic information without requiring direct access to sensitive user data. The deep learning model extracts gender and age characteristics from play behavior patterns, enabling customized marketing while respecting information access restrictions
Solution Approach 2:
The patent changes the parameter space from direct user information (gender, age) to behavioral parameters (playlog data including game completion rates, in-game purchases, social interactions). By transforming the problem from demographic classification to behavioral pattern recognition, the system achieves customization without direct information access
3Measurement precision
If deep learning models are trained using playlog data with account identification information, then prediction accuracy of user information improves, but data processing complexity increases
Solution Approach 1:
The patent performs preliminary actions by pre-processing and labeling playlog data before model training. User information is labeled with account identification information in advance, and playlog data is structured and cleaned beforehand, reducing the computational complexity during the actual prediction phase while maintaining high accuracy
Data Source
AI summary
Disclosed is a method for predicting user information using playlog data performed by a computing device according to some exemplary embodiments of the present disclosure in order to implement the above-mentioned object. The method may include: training a deep learning model through training data to predict the user information from the playlog data, wherein the training data is generated based on first playlog data of account identification information in which the user information is identified; and predicting the user information related to advertisement identification information from second playlog data of the advertisement identification information by the deep learning model.