Automated Feature Engineering Visualization for Predictive Modeling

Resolve Bottlenecks,
Find Innovative SolutionsGenerate Solutions

Solution Overview

Problem

Current machine learning tools for predictive data analytics lack automation in feature engineering, making it a time-consuming and challenging process, especially in relational databases, which is crucial for predictive modeling accuracy.

Innovation Solution

Automating feature engineering by identifying relevant join paths in relational databases, extracting new features, and assigning importance scores to existing and new features using machine learning models, with visualization and explanation capabilities through interactive graphical interfaces and natural language processing.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Manufacturing precision

If manual feature engineering is performed in relational databases, then feature quality and predictive model accuracy can be improved, but time consumption and operational complexity increase significantly

Engineering Contradiction:

Improvepredictive model accuracyVSAvoidtime consumption for feature engineering

Core Design Contradiction:

Manufacturing precisionVSLoss of time

Solution Approach 1:

The system enables automated feature engineering that performs data exploration, join path identification, and feature extraction without requiring manual intervention. The machine learning model automatically evaluates features and generates predictive models, allowing the system to serve itself in the feature engineering process rather than relying on manual analyst input.

Inventive Principle:

Principle #25Self-service

Solution Approach 2:

The system performs preliminary data exploration and automatic identification of relevant join paths before feature extraction. By pre-analyzing the relational database structure and identifying potential feature sources through automated queries, the system prepares the data in advance, reducing the time required for subsequent feature engineering and model building.

Inventive Principle:

Principle #10Preliminary action

2Reliability

If comprehensive feature engineering is performed manually, then feature importance and model reliability can be improved, but operational complexity and difficulty increase

Engineering Contradiction:

Improvemodel reliabilityVSAvoidoperational complexity

Core Design Contradiction:

ReliabilityVSEase of operation

Solution Approach 1:

The system replaces manual mechanical operations of feature engineering with automated computational processes. Machine learning models automatically perform data exploration, evaluate potential features, identify join paths, and generate predictive models, substituting the manual analytical process with algorithmic automation that maintains reliability while reducing operational complexity.

Inventive Principle:

Principle #28Mechanics substitution (Replace mechanical system)

Solution Approach 2:

The system incorporates feedback mechanisms where the machine learning model evaluates feature importance scores and automatically iterates on feature selection. The model provides feedback on which features contribute most to predictive accuracy, allowing automatic refinement of the feature set without requiring manual intervention to assess model performance or adjust features.

Inventive Principle:

Principle #23Feedback

3Productivity

If automated feature engineering is implemented, then productivity and speed can be improved, but the ability to provide detailed explanations and insights may be reduced

Engineering Contradiction:

Improvefeature engineering throughputVSAvoidexplanation quality

Core Design Contradiction:

ProductivityVSLoss of information

Solution Approach 1:

The system introduces an intermediary explanation layer that translates the automated machine learning process into human-understandable insights. The explanation generator acts as a mediator between the automated feature engineering pipeline and the end user, providing detailed information about feature importance, join paths used, and model predictions while maintaining the speed of automated processing.

Inventive Principle:

Principle #24Intermediary (Mediator)

Solution Approach 2:

The system segments the automated feature engineering process into distinct, explainable components: data exploration phase, join path identification phase, feature extraction phase, and model evaluation phase. Each segment can be independently explained and visualized, allowing users to understand the automated process step-by-step without sacrificing overall productivity.

Inventive Principle:

Principle #1Segmentation

Data Source

PatentUS11551123B2Automatic visualization and explanation of feature learning output from a relational database for predictive modelling

Publication Date: 2023.01.10 INTERNATIONAL BUSINESS MACHINE CORPORATION

AI summary

Embodiments for automatic visualization and explanation of feature learning output for predictive modeling in a computing environment by a processor. A degree of importance score may be assigned to one or more features from a relational database according to the machine learning model. A visualization graph of one or more join paths and the one or more features with the degree of importance score to predict a target variable may be generated.