Multidimensional Cube Regression via Unidimensional Chain Indexing
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Traditional multidimensional cube systems are not designed for native machine learning and require data conversion to flat formats, which is computationally intensive and inefficient for fast analytics, making them incompatible with most machine learning algorithms.
Innovation Solution
A method that facilitates compatibility between multivariate regression models and multidimensional datasets by extracting unidimensional chains, double indexing, constructing partial fit regression trees, and repopulating the dataset with predictions, allowing native machine learning operations within multidimensional environments without data conversion.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Adaptability or versatility
If data is converted to flat formats for machine learning, then compatibility with machine learning algorithms is improved, but computational intensity and processing time increase
Solution Approach 1:
The patent segments the multidimensional cube data into unidimensional chains that can be processed independently, allowing machine learning algorithms to operate on structured segments rather than requiring complete flat conversion of the entire dataset
Solution Approach 2:
The patent introduces a dual indexing mechanism as an intermediary layer between the multidimensional cube structure and machine learning algorithms, enabling algorithms to query and process data through the index structure without requiring full data flattening
2Adaptability or versatility
If data is converted to flat formats for machine learning, then compatibility with machine learning algorithms is improved, but analytics speed decreases
Solution Approach 1:
The patent performs preliminary indexing of the multidimensional cube data into unidimensional chains and creates a dual index structure before machine learning operations, so that during actual analytics the pre-indexed structure enables fast queries without repeated data conversion
Solution Approach 2:
The dual index structure serves as an intermediary that maintains the speed benefits of the multidimensional cube structure while enabling machine learning compatibility, allowing fast analytics by querying the index rather than converting entire datasets
3Speed
If traditional multidimensional cube systems are used, then fast analytics are achieved, but compatibility with machine learning algorithms is lost
Solution Approach 1:
The patent makes the multidimensional cube structure universal by adding machine learning capabilities directly to it through the dual indexing mechanism, allowing the same structure to serve both fast analytics and machine learning operations without requiring separate flat data representations
Solution Approach 2:
The dual index acts as an intermediary layer that enables machine learning algorithms to interact with the multidimensional cube structure natively, preserving the speed advantages of the cube structure while adding ML compatibility through the indexing mechanism
Data Source
AI summary
One or more computer processors facilitate compatibility between one or more multivariate regression models and a multidimensional dataset, wherein the program instructions. The one or more computer processors extract a plurality of unidimensional chains from the multidimensional dataset. The one or more computer processors double index the plurality of extracted unidimensional chains. The one or more computer processors construct a plurality of partial fit regression trees from the double indexed unidimensional chains. The one or more computer processors, responsive to a stop criterion, calculate one or more predictions utilizing the plurality of constructed partial fit regression trees. The one or more computer processors repopulate the multidimensional dataset with the one or more calculated predictions.