User-Session Representation Vectors for Topological Data Analysis
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing analysis techniques are ineffective in transforming user interaction data into suitable representation vectors for advanced analysis methods like topological data analysis, particularly due to challenges in encoding categorical data and handling mixed data types with differing scales.
Innovation Solution
A system and method that transform interaction data into user-session representation vectors by normalizing and encoding session datasets, using machine-learning techniques such as recurrent neural networks, to facilitate the application of topological data analysis, preserving the distance between categorical values in a multi-dimensional vector space.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If existing encoding solutions are used to transform interaction data into representation vectors, then the transformation process is simple, but the categorical data cannot be adequately represented in Euclidian space and distances between categorical values cannot be computed
Solution Approach 1:
The patent transforms categorical data parameters into numerical representations by creating embedding vectors that map categorical values to continuous Euclidian space. This allows distance computations between categorical values by changing the parameter representation from discrete categories to continuous vector coordinates, resolving the inability to compute meaningful distances between categorical variables.
Solution Approach 2:
The patent introduces embedding layers as intermediary components that serve as mediators between categorical data and Euclidian space representation. These embedding layers transform categorical inputs into dense vector representations, enabling subsequent distance computations and mathematical operations while maintaining the semantic relationships of the original categorical data.
2Measurement precision
If mixed data types with differing scales are encoded directly, then the encoding process is straightforward, but the distance between values in different categories cannot account for category-specific distance meanings
Solution Approach 1:
The patent applies local quality by creating category-specific embedding spaces where each categorical variable has its own learned distance metrics and scaling characteristics. Instead of applying uniform normalization across all data types, the system allows each category to have its own local transformation rules and distance interpretations, preserving the meaningful differences between categories while enabling coherent distance computations.
3Adaptability or versatility
If interaction data is transformed into representation vectors suitable for advanced techniques like TDA, then advanced analysis can be applied, but existing solutions fail to preserve meaningful distances between categorical values
Solution Approach 1:
The patent performs preliminary action by pre-transforming categorical interaction data into embedding vectors before applying advanced analysis techniques like TDA. This preprocessing step converts categorical data into a format compatible with advanced techniques while preserving meaningful distance relationships, enabling subsequent analysis without losing categorical semantics.
Data Source
AI summary
Disclosed systems and methods generate user-session representation vectors from data generated by user interactions with online services. A transformation application executing on a computing device receives interaction data, which is generated by user devices interacting with an online service. The transformation application separates the interaction data into session datasets. The transformation involves normalizing the session datasets by modifying the rows within each session dataset by removing event identifiers and time stamps. The application transforms each normalized session dataset into a respective user-session representation vector. The application outputs the user-session representation vectors.