Supervised Learning Model Segmentation for Multicollinearity Resolution
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing supervised learning models face challenges in accurately forecasting target variables impacted by both internal and external drivers due to multicollinearity issues, particularly when external drivers have short data histories, leading to unstable forecasts and complex model requirements.
Innovation Solution
The solution involves creating a separate model of external drivers to adjust the target variable, allowing the learning algorithm to focus solely on the adjusted internal factors, thereby eliminating the need for additional forecast models and reducing multicollinearity.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Adaptability or versatility
If supervised learning models combine both internal and external drivers to forecast target variables, then the model comprehensiveness is improved, but multicollinearity between correlated internal and external factors causes instability and complicates the model
Solution Approach 1:
The patent segments the forecasting model into two distinct components: an external driver model that captures time series behavior from external factors, and an internal behavior model that captures idiosyncratic effects. This segmentation separates the correlated external and internal drivers into independent models, eliminating multicollinearity issues while maintaining comprehensive forecasting capability.
2Adaptability or versatility
If external drivers with short data histories are included in the model, then the model's ability to capture time series behavior is improved, but the short history of economic data creates colinearity and causal confusion
Solution Approach 1:
The patent creates a separate external driver model that specifically handles time series behavior from external factors with short histories. This segmentation allows the external model to specialize in capturing temporal patterns from limited economic data, while the internal behavior model focuses on idiosyncratic effects, thereby improving effect separation accuracy.
Solution Approach 2:
The patent introduces an intermediary external driver model that acts as a mediator between external factors and the target variable. This intermediary model processes external drivers separately before their effects are combined with internal behavior, preventing direct colinearity conflicts and causal confusion between short-history external data and longer-history internal data.
3Reliability
If behavioral factors are forecasted as a function of economic factors to resolve correlations, then the correlation problem is addressed, but n additional forecast models must be created, dramatically complicating the task
Solution Approach 1:
Instead of creating n separate forecast models for each behavioral factor as a function of economic factors, the patent segments the system into one external driver model and one internal behavior model. The external driver model captures all time series behavior from external factors, eliminating the need for multiple behavioral factor forecasts and dramatically reducing model complexity.
Data Source
AI summary
A type of machine learning called supervised learning is disclosed. In supervised learning, the training data contains observed values of a target variable and a set of candidate explanatory variables. Supervised learning has been used previously for predicting time series such as economic time series and separately for intrinsic behavior patterns such as credit scoring for offering consumer loans. However, when creating a single model of a target variable that combines both internal behavior and external drivers, the internal behavior and external drivers are often correlated to each other as well as to the target variable. In such a system, the external drivers are usually intended to capture the time series behavior and the internal behavioral variables capture the idiosyncratic effects, but when multicollinearity occurs across all these factors, the internal behavioral variables must also be predicted before a forecast can be created for the target variable. This complicated situation severely limits the interpretability and applicability of such systems.The present invention solves the above described colinearity problem between internal and external factors by first creating a model of how the external factors drive behavior and adjusting the target variable for this known structure prior to creation of the machine learning model. This is similar to the way an offset term is used in generalized linear models (GLM). An external model is used to compute a set of coefficients that are fixed offset during the GLM estimation. The approach provides the same capability to neural network models.This means that the current invention modifies the creation of the model so that the multicollinearity problem is solved such that no time series forecasting of the internal factors is required. All time series structure is concentrated into the initial external model.