Time-Preserving Embeddings for User Activity Sequences
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Service providers face inefficiencies in utilizing machine learning models due to the lack of consideration for temporal dependencies in user activities, leading to inaccurate predictions and redundant feature engineering efforts across multiple service providers.
Innovation Solution
The generation of time-preserving embeddings, which are task-independent and capture sequential and temporal dependencies in user activities, allowing for more accurate model training and reduced computational resources by leveraging a centralized repository of these embeddings.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If service providers perform independent feature engineering for each machine learning task, then task-specific model accuracy may be improved, but computational overhead and processing time increase significantly
Solution Approach 1:
The patent pre-computes and stores embeddings for user trails and activities in advance, creating a centralized repository that can be reused across multiple tasks. This preliminary action eliminates the need to perform feature engineering repeatedly for each task, thereby reducing computational overhead while maintaining prediction accuracy.
Solution Approach 2:
The patent creates universal embeddings that can be applied across multiple different machine learning tasks (recommendations, user segmentation, prediction tasks). These embeddings serve multiple functions simultaneously, eliminating redundant feature engineering efforts across different service providers and tasks.
2Measurement precision
If service providers process large volumes of raw user activity data, then model training accuracy improves, but processing time and computational resources increase
Solution Approach 1:
The patent extracts essential features from large volumes of raw user activity data by creating compressed embeddings that capture the most important patterns in user behavior. This extraction process reduces data dimensionality while preserving the information necessary for accurate model training.
Solution Approach 2:
The patent transforms raw user activity data into a different parameter space through embedding operations. This parameter transformation converts high-dimensional raw data into lower-dimensional vector representations that are more efficient to process while maintaining the essential characteristics needed for accurate predictions.
3Reliability
If service providers engineer features independently for each task, then task-specific performance improves, but redundancy across service providers increases
Solution Approach 1:
The patent merges the feature engineering processes of multiple service providers into a single centralized embedding generation system. By combining efforts to create a shared repository of embeddings, the system eliminates redundant computations across service providers while maintaining the quality needed for various tasks.
Data Source
AI summary
One or more computing devices, systems, and/or methods for generating time-preserving embeddings are provided. User trails of user activities performed by users are generated. Frequencies at which the activities were performed are identified. Indices are assigned to a set of activities identified from the activities as having frequencies above a threshold. Activity descriptions of the set of activities are mapped to the indices to generate a vocabulary. A model is trained using the user trails, timestamps of the activities, and the vocabulary to learn a set of time-preserving embeddings.