Entropy-Based Plotting Space Selection for Data Visualization

Resolve Bottlenecks,
Find Innovative Solutions
Generate Solutions

Solution Overview

Problem

There is a need to automatically determine whether variable data should be plotted in linear or logarithmic space to effectively represent and analyze data, as existing methods require manual trial and error or lack consideration for data distribution characteristics like heteroskedasticity.

Innovation Solution

A computer-assisted method that computes the entropy of data points in both linear and logarithmic spaces, allowing for graphical plotting in the space with lower entropy, which helps in identifying the most appropriate scale for data representation.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Ease of operation

If manual trial and error is used to determine plotting space, then flexibility in analysis is maintained, but time consumption and operational complexity increase

Engineering Contradiction:
Improveease of determining plotting spaceVSAvoidtime for manual trial and error
Core Design Contradiction:
Ease of operationVSLoss of time

Solution Approach 1:

The system automatically determines the optimal plotting space by computing entropy values for both linear and logarithmic transformations and comparing them, eliminating the need for manual trial and error. The method self-selects the appropriate space based on data characteristics, making the system self-sufficient in this analytical decision.

Inventive Principle:
Principle #25Self-service

Solution Approach 2:

The method transforms the data parameter representation by computing entropy in both linear and logarithmic spaces, then selects the transformation that yields lower entropy. This parameter-based approach automates the selection process by evaluating statistical properties rather than relying on manual experimentation.

Inventive Principle:
Principle #35Parameter changes

2Reliability

If logarithmic transformation is applied to financial data, then linear relationships and heteroskedasticity issues are improved, but the transformation process adds computational steps

Engineering Contradiction:
Improvedata representation accuracyVSAvoidcomplexity of transformation process
Core Design Contradiction:
ReliabilityVSDevice complexity

Solution Approach 1:

The method applies logarithmic transformation only when entropy analysis indicates it will improve data representation. By computing entropy for both linear and logarithmic spaces and selecting the better option, the system applies the transformation partially - only when necessary - rather than universally, thus avoiding unnecessary computational complexity while maintaining reliability when the transformation is beneficial.

Inventive Principle:
Principle #16Partial or excessive action

Solution Approach 2:

The system uses entropy computation as feedback to determine whether logarithmic transformation should be applied. By comparing entropy values before and after transformation, the method receives feedback on the quality improvement, enabling intelligent decision-making about whether to proceed with the transformation based on actual data characteristics.

Inventive Principle:
Principle #23Feedback

3Extent of automation

If entropy computation is performed for both linear and logarithmic spaces, then optimal plotting space is automatically selected, but computational resources increase

Engineering Contradiction:
Improveautomation of plotting space selectionVSAvoidcomputational energy consumption
Core Design Contradiction:
Extent of automationVSUse of energy by moving object

Solution Approach 1:

The method performs preliminary entropy computation for both linear and logarithmic transformations before finalizing the plotting space selection. By computing both entropy values in advance and comparing them, the system makes an informed automated decision without requiring iterative adjustments or post-hoc modifications, thus achieving high automation efficiency with a single pass of computations.

Inventive Principle:
Principle #10Preliminary action

Data Source

PatentUS8217944B1Systems and methods for selecting variable transformations
Publication Date: 2012.07.10 MORGAN STANLEY SERVICES GROUP INC
  • US8217944B1 patent drawing
  • US8217944B1 patent drawing
  • US8217944B1 patent drawing

AI summary

A computer-assisted method of determining whether a set of data is to be graphically plotted in linear space or logarithmic space. The method includes receiving a set of data points to be plotted, classifying the data points, and computing a first entropy of a frequency of the data points. The method also includes computing a logarithmic transformation of each of the data points, classifying the logarithmically transformed data points, computing a second entropy of a frequency of the logarithmically transformed data points, and graphically plotting the data points in one of linear space and logarithmic space based on a comparison of the first entropy and the second entropy.