Neural Network Model Merging for User Behavior Prediction
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing methods for predicting user behavior in communications networks face challenges with limited data availability, making it difficult to achieve accurate predictive models, especially in new implementations or regions with data transfer limitations.
Innovation Solution
A node merges pre-existing predictive models from different user groups using an artificial neural network, establishing connections with learned weights based on data from a third group, enabling the creation of a new predictive model with limited data while maintaining accuracy and complexity for the new problem.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If large amounts of data are collected and transferred to train predictive models, then prediction accuracy is improved, but data transfer cost and complexity increase
Solution Approach 1:
The patent segments the predictive model training process by dividing users into different groups (first group with sufficient data, second group with limited data) and training separate predictive models for each group. This segmentation allows the system to leverage sufficient data from the first group while avoiding the need to transfer and process large amounts of data from the second group, thereby maintaining prediction accuracy without incurring high data transfer costs and complexity.
2Measurement precision
If pre-existing predictive models from different user groups are merged using an artificial neural network, then prediction accuracy with limited data is improved, but model complexity increases
Solution Approach 1:
The patent merges pre-existing predictive models from different user groups by establishing connections between them through an artificial neural network. The neural network learns optimal connection weights by processing data from a third group of users, enabling the merged model to achieve high prediction accuracy with limited data. This merging approach consolidates knowledge from multiple groups while using the neural network to manage the complexity of integrating these models.
3Adaptability or versatility
If predictive models are developed for new products or regions with limited data, then adaptability is improved, but prediction accuracy deteriorates
Solution Approach 1:
The patent applies preliminary action by pre-training predictive models on sufficient data from existing user groups (first and second groups) before deploying them to new products or regions. These pre-trained models contain learned patterns and knowledge that can be transferred to new contexts. When data is limited in the new context (third group), the pre-trained models provide a strong foundation, allowing the system to maintain prediction accuracy while adapting to new products or regions.
Data Source
AI summary
A method performed by a node for predicting a behavior of users of a communications network is described. The node manages an artificial neural network. The node merges a first pre-existing predictive model of the behavior in a first group of users with a second model of the behavior in a second group of users. The merging comprises establishing connections between the first model and the second model. Each of the connections has a respective weight. The respective weights of the connections are learned by respective connections of neurons in the artificial neural network based on data from a third group of users. The node also obtains a third model for predicting the behavior in the third group of users, based on the merged models and the data from the third group of users.