Unsupervised Clustering Feature Engineering with Correlation Scoring
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing machine learning algorithms for predicting data outcomes rely on human-generated input features that are limited in breadth and granularity, leading to inefficient and slow training processes.
Innovation Solution
Utilize unsupervised vector quantization algorithms like k-means clustering to automatically partition data records into groupings with correlation scores, generating predictive input features without human intervention, using mutual information scoring to filter out unhelpful data.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Adaptability or versatility
If human-generated input features are used for machine learning algorithms, then the implementation is straightforward and controllable, but the breadth and granularity of features are limited
Solution Approach 1:
The system enables self-service feature engineering by allowing the machine learning algorithm to automatically generate and refine input features through unsupervised clustering, eliminating the need for manual feature creation while improving feature quality and adaptability
Solution Approach 2:
The patent replaces the mechanical process of manual feature engineering with an automated computational system that uses unsupervised clustering algorithms to generate features, transitioning from human-driven to algorithm-driven feature creation
2Reliability
If more comprehensive human-generated features are created to improve prediction accuracy, then the feature quality improves, but the time required for feature generation increases
Solution Approach 1:
The system performs preliminary unsupervised clustering on the training data to pre-generate feature groupings and correlation scores before the actual machine learning training begins, reducing the time required during the main training process while maintaining comprehensive feature coverage
Solution Approach 2:
The patent changes the parameter of feature generation from manual iterative creation to automated algorithmic processing, fundamentally altering the time required by transitioning from hours/days of manual work to minutes of computational processing
3Adaptability or versatility
If unsupervised clustering algorithms are used to automatically generate features, then the feature breadth and granularity improve, but the complexity of the processing system increases
Solution Approach 1:
The patent introduces correlation scoring as an intermediary mechanism that bridges the gap between complex unsupervised clustering results and the machine learning algorithm, providing a simplified metric that indicates the predictive value of each feature grouping
4Reliability
If all generated feature groupings are provided to the machine learning algorithm, then the comprehensiveness of input data improves, but the processing time and computational resources increase
Solution Approach 1:
The patent extracts only the most relevant feature groupings by filtering based on correlation scores, removing unnecessary features from the input set provided to the machine learning algorithm, thus maintaining accuracy while improving training speed
Solution Approach 2:
The system applies partial action by providing only a subset of feature groupings to the machine learning algorithm based on correlation thresholds, rather than providing all possible features, achieving sufficient accuracy with reduced computational overhead
Data Source
AI summary
A method of generating an input for a machine learning algorithm may include collecting data records. Each data record may include a plurality of categories of data. The method may include using vector quantization to partition the plurality of data records into a plurality of groupings. Each of the groupings may be based on one or more of the plurality of categories of data. The method may include generating a correlation score for each of the plurality of groupings. The correlation score may be indicative of whether a particular group is indicative of a given outcome.