Neural Network Execution Blocks for Covariate-Aware Forecasting
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current neural network models for forecasting future values of a target variable suffer from underfitting due to a large number of parameters, especially when the training dataset is limited, and fail to effectively incorporate covariates that significantly influence the target variable.
Innovation Solution
A system and method that utilizes covariate-specific and target-variable-specific AI models to compute the covariate effect on the target variable, remove it from past values, and generate intrinsic forecasts, combining these models iteratively to produce improved forecasts by incorporating covariates such as temporal and categorical factors.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If a large number of neural network parameters are used to model the target variable, then the model can capture complex patterns, but the model underfits the training data when the training dataset size is limited
Solution Approach 1:
The patent segments the forecasting model into two separate neural networks: a covariate-specific AI model that processes covariate inputs and computes their effects on the target variable, and a target-variable-specific AI model that processes the target variable's past values. This segmentation reduces the parameter count in each individual model while maintaining the ability to capture complex patterns through the interaction of the two models.
Solution Approach 2:
The patent extracts and isolates the covariate effects from the target variable forecasting process. The covariate-specific AI model separately computes the covariate effects, which are then integrated with the target-variable-specific AI model's predictions. This extraction allows each model to focus on specific aspects of the forecasting problem, reducing overall complexity.
2Adaptability or versatility
If traditional neural network models are used for forecasting, then the model structure is simple, but the model fails to effectively incorporate covariates that significantly influence the target variable
Solution Approach 1:
The patent implements segmentation by creating distinct processing pathways: one for covariates through the covariate-specific AI model and another for the target variable through the target-variable-specific AI model. This allows the system to effectively incorporate multiple types of inputs while maintaining clear functional separation.
Solution Approach 2:
The patent achieves multi-functionality by designing a hybrid architecture where the covariate-specific AI model handles external influencing factors and the target-variable-specific AI model handles the temporal dynamics of the target variable. Together, they provide a universal forecasting solution that accommodates both covariate effects and time-series patterns.
Data Source
AI summary
A method for forecasting future values of a target variable using past values thereof, the values of the target variable being affected by one or more covariates wherein the covariates are independent from the target variable. The method comprises using a covariate-specific AI model, computing a covariate effect of the covariates on the target variable. The covariates effect is a defined modification to the values of the target variable caused by the covariates. The method also comprises computing intrinsic past values of the target variable by removing the covariate effect of the covariates from past values of the target variable. The method further comprises using a target-variable-specific AI model, generating an intrinsic forecast of the future values of the target variable; and computing a forecast that includes the covariate effect using the intrinsic forecast of the future values of the target variable and the covariate effect.