Predictive Model Generation via Automated Variable Transformation
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Predictive modeling in commercial settings, such as targeted marketing, often relies on analyst expertise and is hindered by the time and expense of creating models, with inaccuracies stemming from data quality and tool performance, particularly when dealing with large datasets containing redundant or irrelevant attributes.
Innovation Solution
A machine-based method that enables users to automatically generate predictive models by transforming variables and ranking predictor variables, providing graphical displays to identify distinguishing customer segments, and facilitating efficient model development through a project-based workflow with automated data preparation and validation.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If manual model development by analysts is used, then model accuracy can be achieved through expertise, but development time and cost increase significantly
Solution Approach 1:
The system enables self-service model development where the software automatically performs variable transformation, model generation, and validation without requiring analyst intervention for these routine tasks. The automated workflow allows users to simply load data and select objectives while the system handles the complex computational processes independently.
Solution Approach 2:
The patent replaces the manual mechanical process of model building with an automated computational system. Instead of analysts manually selecting variables, transforming data, and iteratively building models, the system uses algorithms to automatically perform these operations, substituting human expertise with computational automation.
2Manufacturing precision
If analysts manually transform and select variables, then model quality can be controlled, but complexity and redundancy in data processing increase
Solution Approach 1:
The system segments the data processing workflow into distinct automated stages: data loading, variable transformation, model generation, validation, and deployment. Each stage is handled by specialized computational modules that process specific aspects of the data independently, reducing overall complexity while maintaining quality control through structured processing.
Solution Approach 2:
The system automatically applies parameter transformations to raw data, converting variables into optimized forms suitable for modeling. This includes automatic handling of missing values, creation of dummy variables for categorical data, and transformation of continuous variables, all performed through standardized parameter changes rather than manual processing.
3Productivity
If automated variable transformation is applied, then development time is reduced, but control over model decisions may be lost
Solution Approach 1:
The system incorporates feedback mechanisms that allow analysts to review automated decisions and adjust parameters as needed. After automatic model generation, the system provides summaries of variable transformations and model results, enabling analysts to verify outcomes and make corrective adjustments if necessary, thus maintaining control while enjoying automation benefits.
Solution Approach 2:
The system provides dynamic control where analysts can adjust model parameters, selection criteria, and transformation options based on their specific needs. The automated workflow adapts to user inputs and preferences, allowing flexible control over the modeling process while maintaining high productivity through automation of routine operations.
Data Source
AI summary
Models are generated using a variety of tools and features of a model generation platform. For example, in connection with a project in which a user generates a predictive model based on historical data about a system being modeled, the user is provided through a graphical user interface a structured sequence of model generation activities to be followed, the sequence including dimension reduction, model generation, model process validation, and model re-generation.In connection with a process in which a user generates a collection of predictive models or an aggregate predictive model based on historical data about a system being modeled, profiles of aggregate targets are generated based on key contributory variables.