Image Representation of Categorical Data for Predictive Analysis
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing predictive data analysis systems are inefficient and unreliable in handling high-dimensional categorical feature spaces with high cardinality, making them unsuitable for complex input structures.
Innovation Solution
The use of image transformations, such as reserved-spatial-location, coordinate-based, and feature-based transformations, to generate image representations of categorical input features, which can be processed by image-based machine learning models like CNNs for efficient and reliable predictive data analysis.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Productivity
If existing predictive data analysis systems are used to handle high-dimensional categorical feature spaces, then the systems can process the data, but the computational efficiency and prediction reliability deteriorate
Solution Approach 1:
The patent introduces an image transformation intermediary that converts high-dimensional categorical feature data into image representations. This intermediary layer enables existing CNN architectures to process categorical data effectively, resolving the contradiction between computational efficiency and prediction reliability by bridging the gap between data types and model requirements
Solution Approach 2:
The patent transforms the parameter representation of categorical features by converting them into spatial coordinates and image pixel values. This parameter transformation allows the data to be processed by image-based models, simultaneously improving computational efficiency through leveraged CNN optimizations and enhancing prediction reliability through appropriate feature representation
2Productivity
If traditional predictive analysis methods are applied to complex input structures, then the methods can operate, but the training efficiency and accuracy deteriorate
Solution Approach 1:
The patent applies dimensionality change by transforming categorical feature data into a two-dimensional image space. This allows the data to be processed by CNNs that are optimized for image data, thereby improving training efficiency while maintaining or enhancing prediction accuracy through the spatial relationships preserved in the image representation
3Reliability
If image transformations are applied to categorical feature data, then training efficiency and accuracy improve, but the system complexity increases
Solution Approach 1:
The patent segments the feature processing pipeline into distinct stages: coordinate generation from categorical features, image representation construction, and CNN processing. This segmentation allows each component to be optimized independently and reused across different applications, managing system complexity while maintaining improved prediction reliability
Data Source
AI summary
There is a need for more effective and efficient predictive data analysis solutions and/or more effective and efficient solutions for generating image representations of categorical/scalar data. Various embodiments of the present invention address one or more of the noted technical challenges. In one example, a method comprises receiving the one or more categorical input features; generating an image representation of the one or more categorical input features, wherein the image representation comprises image region values each associated with a categorical input feature, and further wherein each image region value of the one or more image region values is determined based at least in part on the corresponding categorical input feature associated with the image region value; and processing the image representation using an image-based machine learning model to generate the image-based predictions.