Bayesian Network Data Fusion for High-Dimensional Insights

Resolve Bottlenecks,
Find Innovative Solutions
Generate Solutions

Solution Overview

Problem

Fusing and visualizing diverse datasets from disparate sources with different granularities and lacking natural join keys poses challenges in deriving insights, especially when dealing with high-dimensional data, where traditional query mechanisms are inefficient and inaccurate.

Innovation Solution

The system learns a Bayesian network using a minimum spanning tree embedded in a graph derived from pair-wise mutual information values to predict a target attribute, enabling efficient data fusion and visualization through exact inference and conditional probabilistic tables, while also providing data completion and probabilistic joins across datasets.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Measurement precision

If traditional query mechanisms are used to derive insights from high-dimensional fused data, then the data can be processed, but the efficiency and accuracy deteriorate

Engineering Contradiction:
Improveaccuracy of insight derivationVSAvoidefficiency of data processing
Core Design Contradiction:
Measurement precisionVSProductivity

Solution Approach 1:

The patent introduces Bayesian networks as an intermediary mechanism between raw fused data and insight derivation. The Bayesian network learns probabilistic relationships from the data and enables efficient querying through conditional probability calculations, avoiding the inefficiency of traditional queries on high-dimensional data while maintaining accuracy through probabilistic inference

Inventive Principle:
Principle #24Intermediary (Mediator)

Solution Approach 2:

The patent transforms the data representation by learning a Bayesian network structure and parameters from the fused data. This parameter transformation converts raw high-dimensional data into a compact probabilistic model with learned conditional probability tables, enabling efficient and accurate queries through the model rather than direct data processing

Inventive Principle:
Principle #35Parameter changes

2Quantity of substance

If datasets from disparate sources with different granularities are fused, then more information is available, but the complexity of data integration increases

Engineering Contradiction:
Improveamount of informationVSAvoidcomplexity of data integration
Core Design Contradiction:
Quantity of substanceVSDevice complexity

Solution Approach 1:

The patent applies a universal Bayesian network framework that can handle diverse data types and granularities through a common probabilistic modeling approach. The Bayesian network structure automatically adapts to different data sources by learning appropriate relationships, providing a unified mechanism for integrating heterogeneous datasets without requiring source-specific processing logic

Inventive Principle:
Principle #6Universality (Multi-functionality)

Solution Approach 2:

The Bayesian network learning process automatically discovers the data structure and relationships from the fused datasets without manual intervention. The system self-configures the network structure and conditional probability tables by learning from the data itself, eliminating the need for complex manual data integration configurations

Inventive Principle:
Principle #25Self-service

Data Source

PatentEP3217334B1System and method for visual bayesian data fusion
Publication Date: 2021.09.29 TATA CONSULTANCY SERVICES LTD
  • EP3217334B1 patent drawingFigure 1
  • EP3217334B1 patent drawingFigure 2
  • EP3217334B1 patent drawingFigure 3A

AI summary

System and method for visual Bayesian data fusion are disclosed. In an example, a plurality of datasets associated with a topic are obtained from a data lake. Each of the plurality of datasets include information corresponding to various attributes of the topic. Further, the plurality of datasets are joined to obtain a joined dataset. Furthermore, distribution associated with a target attribute is predicted using Bayesian modeling by selecting a plurality of attributes (k) based on mutual information with the target attribute in the joined dataset, learning a minimum spanning tree based Bayesian structure using the selected attributes and the target attribute, learning conditional probabilistic tables at each node of the minimum spanning tree based Bayesian structure; and predicting the distribution associated with the target attribute by querying the conditional probabilistic tables, thereby facilitating visual Bayesian data fusion.